Ideone.com

fork download

copy

import math
import random
import time
 
 
def play(gameState, memo={}):
    start_time = time.process_time()
    turn = gameState.currentTurn
    maxT = gameState.maxTurn
 
    INF = 10 ** 9
    WIN = 100000
    PLYLIM = 44
 
    W = gameState.width
    H = gameState.height
    goal_mode = gameState.goalMode
 
    walls_xy = frozenset((w.x, w.y) for w in gameState.walls)
    goals0_xy = frozenset((g.x, g.y) for g in gameState.goals) or frozenset([(W, H)])
    goals1_xy = frozenset((g.x, g.y) for g in gameState.enemyGoals) or frozenset([(1, 1)])
 
    # Bảng khắc chế: 0=R (Búa), 1=P (Bao), 2=S (Kéo)
    BT = ((False, False, True), (True, False, False), (False, True, False))
    PRED = (1, 2, 0)
    PREY = (2, 0, 1)
 
    ELIM1 = 65
    ELIM2 = 0
 
    # Nếu trận trước bị reset giữa chừng thì xóa cache cũ
    if turn <= 1 and 'tabs' in memo:
        memo.clear()
 
    map_sig = (W, H, walls_xy, goals0_xy, goals1_xy, goal_mode)
    tabs = memo.get('tabs')
 
    # Khởi tạo bảng tra cứu bản đồ (chạy 1 lần ở lượt đầu hoặc khi đổi map)
    if tabs is None or memo.get('map_sig') != map_sig:
        memo['map_sig'] = map_sig
        memo['tt'] = {}
        memo['hist'] = None
        memo['seen'] = {}
        memo['cfg'] = {}
 
        STRIDE = W + 2
        SIZE = (W + 2) * (H + 2)
        OFFS = (STRIDE + 1, STRIDE, 1, STRIDE - 1, -(STRIDE - 1), -1, -STRIDE, -(STRIDE + 1))
        DIRN = {
            STRIDE - 1: 1,
            STRIDE: 2,
            STRIDE + 1: 3,
            1: 4,
            -(STRIDE - 1): 5,
            -STRIDE: 6,
            -(STRIDE + 1): 7,
            -1: 8,
        }
 
        wall_set = frozenset(
            y * STRIDE + x for (x, y) in walls_xy if 1 <= x <= W and 1 <= y <= H
        )
        cells = [
            y * STRIDE + x
            for y in range(1, H + 1)
            for x in range(1, W + 1)
            if (y * STRIDE + x) not in wall_set
        ]
        valid_set = set(cells)
 
        GOAL0 = frozenset(y * STRIDE + x for (x, y) in goals0_xy if (y * STRIDE + x) in valid_set)
        GOAL1 = frozenset(y * STRIDE + x for (x, y) in goals1_xy if (y * STRIDE + x) in valid_set)
        GOAL = (GOAL0, GOAL1)
 
        nb0_set = {g + off for g in GOAL0 for off in OFFS if (g + off) in valid_set and (g + off) not in GOAL0}
        nb1_set = {g + off for g in GOAL1 for off in OFFS if (g + off) in valid_set and (g + off) not in GOAL1}
        NB = (tuple(nb0_set), tuple(nb1_set))
        DEF_SQ = (GOAL0 | nb0_set, GOAL1 | nb1_set)
 
        MAX_D = W + H + 12
 
        # BFS tính khoảng cách ngắn nhất né tường tới tập ô đích của ta (D0) và địch (D1)
        D0 = [MAX_D] * SIZE
        bfs_q = list(GOAL0)
        for g in GOAL0:
            D0[g] = 0
        head = 0
        while head < len(bfs_q):
            u = bfs_q[head]
            head += 1
            du = D0[u] + 1
            for off in OFFS:
                v = u + off
                if v in valid_set and du < D0[v]:
                    D0[v] = du
                    bfs_q.append(v)
 
        D1 = [MAX_D] * SIZE
        bfs_q = list(GOAL1)
        for g in GOAL1:
            D1[g] = 0
        head = 0
        while head < len(bfs_q):
            u = bfs_q[head]
            head += 1
            du = D1[u] + 1
            for off in OFFS:
                v = u + off
                if v in valid_set and du < D1[v]:
                    D1[v] = du
                    bfs_q.append(v)
 
        # BFS tính khoảng cách thực tế giữa mọi cặp ô hợp lệ để săn đuổi vòng qua vật cản
        DT = [[MAX_D] * SIZE for _ in range(SIZE)]
        MDT = [[MAX_D * 2] * SIZE for _ in range(SIZE)]
        for src in cells:
            row = DT[src]
            mrow = MDT[src]
            row[src] = 0
            mrow[src] = 0
            bfs_q = [src]
            head = 0
            while head < len(bfs_q):
                u = bfs_q[head]
                head += 1
                du = row[u] + 1
                for off in OFFS:
                    v = u + off
                    if v in valid_set and du < row[v]:
                        row[v] = du
                        mrow[v] = du * 2
                        bfs_q.append(v)
 
        max_d0 = max((D0[a] for a in cells if D0[a] < MAX_D), default=1)
        max_d1 = max((D1[a] for a in cells if D1[a] < MAX_D), default=1)
        ADV0 = [0] * SIZE
        ADV1 = [0] * SIZE
        CENTER = [0] * SIZE
        mid_x = (W + 1) * 0.5
        mid_y = (H + 1) * 0.5
        for a in cells:
            x = a % STRIDE
            y = a // STRIDE
            ADV0[a] = max(0, max_d0 - D0[a])
            ADV1[a] = max(0, max_d1 - D1[a])
            cx = abs(x - mid_x) / max(1.0, W / 9.0)
            cy = abs(y - mid_y) / max(1.0, H / 9.0)
            CENTER[a] = max(0, int(4 - max(cx, cy)))
 
        # Bảng nước đi hợp lệ sắp xếp sẵn theo độ giảm khoảng cách tới đích
        MTAB0 = [() for _ in range(SIZE)]
        MTAB1 = [() for _ in range(SIZE)]
        for a in cells:
            m0 = []
            m1 = []
            for off in OFFS:
                q = a + off
                if q in valid_set:
                    h_ix = a * SIZE + q
                    fw0_b = (D0[a] - D0[q]) * 2
                    fw1_b = (D1[a] - D1[q]) * 2
                    m0.append((q, fw0_b, h_ix, 1 if q in nb0_set else 0))
                    m1.append((q, fw1_b, h_ix, 1 if q in nb1_set else 0))
            m0.sort(key=lambda item: item[1], reverse=True)
            m1.sort(key=lambda item: item[1], reverse=True)
            MTAB0[a] = tuple(m0)
            MTAB1[a] = tuple(m1)
        MTAB = (tuple(MTAB0), tuple(MTAB1))
 
        # Bảng điểm thưởng khoảng cách tới đích (tự co giãn theo kích thước map)
        RT = tuple([0] + [int(950 * (0.60 ** (d - 1))) for d in range(1, MAX_D + 1)])
        UNSTOP_RT = tuple([0] + [int(1750 * (0.64 ** (d - 1))) for d in range(1, MAX_D + 1)])
 
        rz = random.Random(20260402)
        Z = tuple(tuple(rz.getrandbits(52) for _ in range(SIZE)) for _ in range(48))
        ZS = rz.getrandbits(52)
 
        tabs = (
            STRIDE, SIZE, MAX_D, OFFS, DIRN, wall_set, GOAL, NB, DEF_SQ,
            DT, MDT, ADV0, ADV1, D0, D1, CENTER, MTAB, RT, UNSTOP_RT, Z, ZS
        )
        memo['tabs'] = tabs
 
    (
        STRIDE, SIZE, MAX_D, OFFS, DIRN, wall_set, GOAL, NB, DEF_SQ,
        DT, MDT, ADV0, ADV1, D0, D1, CENTER, MTAB, RT, UNSTOP_RT, Z, ZS
    ) = tabs
 
    # Quản lý ngân sách CPU: tổng 180s/ván, tối đa 1s/nước (chốt ở 0.82s để an toàn)
    used = memo.get('used', 0.0)
    left = max(1, (maxT - turn) // 2 + 1)
    if turn <= 1:
        budget = 0.82
    else:
        budget = min(0.82, max(0.04, (160.0 - used) / min(left, 150)))
    deadline = start_time + budget
 
    # Dựng bàn cờ 1 chiều: ngoài biên và ô tường đen mang giá trị -1
    board = [-1] * SIZE
    for yy in range(1, H + 1):
        base_y = yy * STRIDE
        for xx in range(1, W + 1):
            pos = base_y + xx
            if pos not in wall_set:
                board[pos] = 0
 
    # Đánh chỉ số động cho quân hai bên (hỗ trợ map có số quân khác 9)
    ally_list = list(gameState.ally)
    opp_list = list(gameState.opponent)
    na_init = len(ally_list)
    nb_init = len(opp_list)
 
    ALLY_LO = 1
    ALLY_HI = na_init
    OPP_LO = na_init + 2
    OPP_HI = OPP_LO + nb_init - 1
    LO = (ALLY_LO, OPP_LO)
    HI = (ALLY_HI, OPP_HI)
 
    slot_cap = max(48, OPP_HI + 2)
    if len(Z) < slot_cap:
        rz = random.Random(20260402)
        Z = tuple(tuple(rz.getrandbits(52) for _ in range(SIZE)) for _ in range(slot_cap))
 
    tp = [0] * slot_cap
    ps = [-1] * slot_cap
    pid_of = [0] * slot_cap
    TI = {'R': 0, 'P': 1, 'S': 2}
    cnt = [0, 0]
 
    for idx_a, p in enumerate(ally_list):
        c = ALLY_LO + idx_a
        pid_of[c] = p.pid
        tp[c] = TI[p.Type]
        pos = p.y * STRIDE + p.x
        ps[c] = pos
        board[pos] = c
        cnt[0] += 1
 
    for idx_b, p in enumerate(opp_list):
        c = OPP_LO + idx_b
        pid_of[c] = p.pid
        tp[c] = TI[p.Type]
        pos = p.y * STRIDE + p.x
        ps[c] = pos
        board[pos] = c
        cnt[1] += 1
 
    tt = memo.get('tt')
    if tt is None or len(tt) > 120000:
        tt = {}
        memo['tt'] = tt
 
    hist_len = SIZE * SIZE
    HIST = memo.get('hist')
    if HIST is None or len(HIST) != hist_len:
        HIST = [0] * hist_len
        memo['hist'] = HIST
    else:
        for i in range(hist_len):
            if HIST[i]:
                HIST[i] >>= 1
 
    KILLER = [None] * 64
    nodes = [0]
    abort = [False]
 
    total = cnt[0] + cnt[1]
    if memo.get('prev_total') != total:
        memo['prev_total'] = total
        memo['capt_turn'] = turn
        memo['cfg'] = {}
    stag = (turn - memo.get('capt_turn', turn)) // 2
    ahead = cnt[0] > cnt[1]
    behind = cnt[0] < cnt[1]
 
    K = 5 + min(8, turn // 25)
    if not behind and stag > 16:
        K += min(7, (stag - 16) // 6)
    if behind:
        HW = 3
    elif stag > 16:
        HW = 11
    else:
        HW = 6
    if ahead and stag > 26:
        DP = (0, 220, 120, 55)
    else:
        DP = (0, 330, 175, 80)
 
    # Phân vai chiến thuật: DEFENDER (chặn đích nhà), RUNNER (đột phá), HUNTER (săn quân)
    roles = {}
    e_types_alive = [0, 0, 0]
    e_near_home = []
    e_near_away = [0, 0, 0]
    for c in range(OPP_LO, OPP_HI + 1):
        q = ps[c]
        if q >= 0:
            u = tp[c]
            e_types_alive[u] += 1
            if D1[q] <= 4:
                e_near_home.append((D1[q], q, u))
            if D0[q] <= 2:
                e_near_away[u] += 1
    e_near_home.sort()
 
    def_assigned = set()
    for d_h, q_h, u_h in e_near_home:
        need_t = PRED[u_h]
        best_c = 0
        best_d = 999
        for c in range(ALLY_LO, ALLY_HI + 1):
            p = ps[c]
            if p >= 0 and tp[c] == need_t and c not in def_assigned:
                d_c = D1[p]
                if d_c < best_d:
                    best_d = d_c
                    best_c = c
        if best_c:
            roles[best_c] = 'DEFENDER'
            def_assigned.add(best_c)
 
    if not def_assigned:
        best_c = 0
        best_d = 999
        for c in range(ALLY_LO, ALLY_HI + 1):
            p = ps[c]
            if p >= 0 and e_types_alive[PREY[tp[c]]] > 0:
                d_c = D1[p]
                if d_c < best_d:
                    best_d = d_c
                    best_c = c
        if best_c and best_d <= 3:
            roles[best_c] = 'DEFENDER'
            def_assigned.add(best_c)
 
    runner_cands = []
    for c in range(ALLY_LO, ALLY_HI + 1):
        p = ps[c]
        if p >= 0 and c not in def_assigned:
            t = tp[c]
            imm = 1 if e_types_alive[PRED[t]] == 0 else 0
            brk = 1 if e_near_away[PREY[t]] > 0 else 0
            runner_cands.append((D0[p] - 2 * imm - brk, -ADV0[p], c))
    runner_cands.sort()
    num_runners = max(2, len(GOAL[0])) if goal_mode == 1 else 2
    for idx_r in range(min(num_runners, len(runner_cands))):
        roles[runner_cands[idx_r][2]] = 'RUNNER'
    for c in range(ALLY_LO, ALLY_HI + 1):
        if ps[c] >= 0 and c not in roles:
            roles[c] = 'HUNTER'
 
    # Kiểm tra phe s có thể chạm đích thắng ngay trong 1 nước đi không
    def wins_now(s):
        lo = LO[s]
        hi = HI[s]
        gset = GOAL[s]
        if goal_mode == 0:
            for g in gset:
                v0 = board[g]
                if lo <= v0 <= hi:
                    return True
                if v0 == 0:
                    for off in OFFS:
                        v = board[g + off]
                        if lo <= v <= hi:
                            return True
                else:
                    t = tp[v0]
                    for off in OFFS:
                        v = board[g + off]
                        if lo <= v <= hi and BT[tp[v]][t]:
                            return True
            return False
        else:
            missing = [g for g in gset if not (lo <= board[g] <= hi)]
            if not missing:
                return True
            if len(missing) > 1:
                return False
            g = missing[0]
            v0 = board[g]
            for off in OFFS:
                src = g + off
                v = board[src]
                if lo <= v <= hi and src not in gset:
                    if v0 == 0 or BT[tp[v]][tp[v0]]:
                        return True
            return False
 
    # Sinh nước đi hợp lệ: mode=0 (tất cả), mode=1 (chỉ ăn quân), mode=2 (ăn quân + áp sát đích)
    def gen(s, mode):
        res = []
        lo = LO[s]
        hi = HI[s]
        elo = LO[1 - s]
        ehi = HI[1 - s]
        mtab = MTAB[s]
        gset = GOAL[s]
        egset = GOAL[1 - s]
        for c in range(lo, hi + 1):
            p = ps[c]
            if p < 0:
                continue
            t = tp[c]
            bt = BT[t]
            on_goal = (goal_mode == 1 and p in gset)
            for q, fw_b, h_ix, is_nb in mtab[p]:
                v = board[q]
                if v == 0:
                    if mode == 0:
                        h = HIST[h_ix]
                        pen_leave = -35 if (on_goal and q not in gset) else 0
                        res.append((fw_b + pen_leave + (h if h < 72 else 72), c, p, q, 0))
                    elif mode == 2 and is_nb:
                        for off in OFFS:
                            gq = q + off
                            if gq in gset:
                                vg = board[gq]
                                if vg == 0 or (elo <= vg <= ehi and bt[tp[vg]]):
                                    res.append((820 + fw_b, c, p, q, 0))
                                    break
                elif elo <= v <= ehi:
                    if bt[tp[v]]:
                        bonus = 90 if (q in egset or D1[q] <= 2) else 0
                        res.append((1050 + fw_b + bonus, c, p, q, v))
        return res
 
    # Hàm lượng giá thế cờ tĩnh từ góc nhìn phe s
    def evaluate_raw(s):
        A_p = []
        A_t = []
        B_p = []
        B_t = []
        cA0 = cA1 = cA2 = 0
        cB0 = cB1 = cB2 = 0
        for c in range(ALLY_LO, ALLY_HI + 1):
            p = ps[c]
            if p >= 0:
                t = tp[c]
                A_p.append(p)
                A_t.append(t)
                if t == 0:
                    cA0 += 1
                elif t == 1:
                    cA1 += 1
                else:
                    cA2 += 1
        for c in range(OPP_LO, OPP_HI + 1):
            p = ps[c]
            if p >= 0:
                t = tp[c]
                B_p.append(p)
                B_t.append(t)
                if t == 0:
                    cB0 += 1
                elif t == 1:
                    cB1 += 1
                else:
                    cB2 += 1
 
        na = len(A_p)
        nb = len(B_p)
        if na == 0:
            return -WIN + 100 if s == 0 else WIN - 100
        if nb == 0:
            return WIN - 100 if s == 0 else -WIN + 100
 
        cA = (cA0, cA1, cA2)
        cB = (cB0, cB1, cB2)
 
        # Hơn quân và lợi thế tuyệt đối khi địch mất sạch loại quân khắc chế
        v = 1050 * (na - nb) + 3200 * (na - nb) // (na + nb)
 
        for t in (0, 1, 2):
            pred = PRED[t]
            prey = PREY[t]
            ca = cA[t]
            cb = cB[t]
            if ca:
                if cB[pred] == 0:
                    v += ca * (500 if cB[t] == 0 else 360)
                elif cB[pred] == 1 and cA[prey] > 0:
                    v += ca * ELIM1
                elif ELIM2 and cB[pred] == 2 and cA[prey] >= 2:
                    v += ca * ELIM2
                v += ca * (80 * cB[prey] if cB[prey] else -50)
            if cb:
                if cA[pred] == 0:
                    v -= cb * (500 if cA[t] == 0 else 360)
                elif cA[pred] == 1 and cB[prey] > 0:
                    v -= cb * ELIM1
                elif ELIM2 and cA[pred] == 2 and cB[prey] >= 2:
                    v -= cb * ELIM2
                v -= cb * (80 * cA[prey] if cA[prey] else -50)
 
        # Điểm chiếm đóng đồng thời khi bản đồ bật goalMode == 1
        if goal_mode == 1:
            occ0 = 0
            for g in GOAL[0]:
                if ALLY_LO <= board[g] <= ALLY_HI:
                    occ0 += 1
            occ1 = 0
            for g in GOAL[1]:
                if OPP_LO <= board[g] <= OPP_HI:
                    occ1 += 1
            if occ0 == len(GOAL[0]):
                return WIN - 50 if s == 0 else -WIN + 50
            if occ1 == len(GOAL[1]):
                return -WIN + 50 if s == 0 else WIN - 50
            v += 900 * (occ0 - occ1)
 
        ha = [MAX_D] * na
        pa = [MAX_D] * na
        ma = [MAX_D * 2] * na
        inta = [999] * na
        hb = [MAX_D] * nb
        pb = [MAX_D] * nb
        mb = [MAX_D * 2] * nb
        intb = [999] * nb
 
        for i in range(na):
            p = A_p[i]
            t = A_t[i]
            d0p = D0[p]
            row = DT[p]
            mrow = MDT[p]
            bt = BT[t]
            for j in range(nb):
                q = B_p[j]
                u = B_t[j]
                d0q = D0[q]
                d1q = D1[q]
                if u == t:
                    if d0q <= d0p and d0q < inta[i]:
                        inta[i] = d0q
                    if D1[p] <= d1q and D1[p] < intb[j]:
                        intb[j] = D1[p]
                    continue
                d = row[q]
                md = mrow[q]
                if bt[u]:
                    if d < pa[i]:
                        pa[i] = d
                    if md < ma[i]:
                        ma[i] = md
                    if d < hb[j]:
                        hb[j] = d
                    if D1[p] <= d1q + 1 and D1[p] < intb[j]:
                        intb[j] = D1[p]
                else:
                    if d < ha[i]:
                        ha[i] = d
                    if d < pb[j]:
                        pb[j] = d
                    if md < mb[j]:
                        mb[j] = md
                    if d0q <= d0p + 1 and d0q < inta[i]:
                        inta[i] = d0q
 
        adv0 = 0
        thr0 = 0
        dg0 = 0
        hunt0 = 0
        hang0 = 0
        away_guard_b = [0, 0, 0]
        for j in range(nb):
            if D0[B_p[j]] <= 2:
                away_guard_b[B_t[j]] = 1
 
        max_unstop = len(UNSTOP_RT) - 1
        max_rt = len(RT) - 1
        hunt_range = max(10, (W + H) // 2 + 2)
 
        for i in range(na):
            p = A_p[i]
            t = A_t[i]
            ap = ADV0[p]
            if cB[PRED[t]] == 0:
                ap += ap >> 1
            if away_guard_b[PREY[t]]:
                ap += max(0, hunt_range - D0[p]) * 3
            adv0 += ap
 
            mp = pa[i]
            if mp < MAX_D:
                hunt0 += HW * max(0, hunt_range - mp) + max(0, hunt_range * 2 - ma[i])
            else:
                hunt0 += CENTER[p]
 
            mh = ha[i]
            if mh <= 3:
                pt = PREY[t]
                row = DT[p]
                prot = False
                for k in range(na):
                    if A_t[k] == pt and row[A_p[k]] <= 1:
                        prot = True
                        break
                if not prot:
                    if mh <= 1:
                        dg0 += DP[1] + (160 if s == 1 else 0)
                        hang0 += 1
                    else:
                        dg0 += DP[mh]
                elif mh <= 2:
                    hunt0 += 25
 
            d = D0[p]
            if inta[i] == 999:
                if d <= max_unstop:
                    tv = UNSTOP_RT[d]
                    if tv > thr0:
                        thr0 = tv
            elif d <= max_rt:
                tv = RT[d]
                if mh <= 2 or inta[i] <= d:
                    tv //= 3
                if tv > thr0:
                    thr0 = tv
 
        if hang0 >= 2:
            dg0 += 380 * (hang0 - 1)
 
        adv1 = 0
        thr1 = 0
        dg1 = 0
        hunt1 = 0
        hang1 = 0
        home_guard_a = [0, 0, 0]
        for i in range(na):
            if D1[A_p[i]] <= 2:
                home_guard_a[A_t[i]] = 1
 
        for j in range(nb):
            q = B_p[j]
            u = B_t[j]
            aq = ADV1[q]
            if cA[PRED[u]] == 0:
                aq += aq >> 1
            if home_guard_a[PREY[u]]:
                aq += max(0, hunt_range - D1[q]) * 3
            adv1 += aq
 
            mp = pb[j]
            if mp < MAX_D:
                hunt1 += HW * max(0, hunt_range - mp) + max(0, hunt_range * 2 - mb[j])
            else:
                hunt1 += CENTER[q]
 
            mh = hb[j]
            if mh <= 3:
                pt = PREY[u]
                row = DT[q]
                prot = False
                for k in range(nb):
                    if B_t[k] == pt and row[B_p[k]] <= 1:
                        prot = True
                        break
                if not prot:
                    if mh <= 1:
                        dg1 += DP[1] + (160 if s == 0 else 0)
                        hang1 += 1
                    else:
                        dg1 += DP[mh]
                elif mh <= 2:
                    hunt1 += 25
 
            d = D1[q]
            if intb[j] == 999:
                if d <= max_unstop:
                    tv = UNSTOP_RT[d]
                    if tv > thr1:
                        thr1 = tv
            elif d <= max_rt:
                tv = RT[d]
                if mh <= 2 or intb[j] <= d:
                    tv //= 3
                if tv > thr1:
                    thr1 = tv
 
        if hang1 >= 2:
            dg1 += 380 * (hang1 - 1)
 
        v += K * (adv0 - adv1) + (thr0 - thr1) - (dg0 - dg1) + (hunt0 - hunt1)
 
        # Phòng thủ vùng cờ nhà
        has_useful_def = False
        for i in range(na):
            if D1[A_p[i]] <= 2 and cB[PREY[A_t[i]]] > 0:
                has_useful_def = True
                break
 
        near_cnt = 0
        for j in range(nb):
            d = D1[B_p[j]]
            if d <= 2:
                v -= 520 * (3 - d)
                near_cnt += 1
                if intb[j] == 999:
                    v -= 450 * (3 - d)
            elif d == 3:
                v -= 110
                if intb[j] == 999:
                    v -= 220
            elif d == 4:
                v -= 35
 
        if near_cnt >= 2:
            v -= 420 * (near_cnt - 1)
        if near_cnt >= 1 and not has_useful_def:
            v -= 420 * near_cnt
        if has_useful_def:
            v += 135 + 105 * near_cnt
 
        return v if s == 0 else -v
 
    evc = {}
 
    def evaluate(s, h):
        r = evc.get(h)
        if r is None:
            r = evaluate_raw(s)
            evc[h] = r
        return r
 
    # Tìm kiếm tĩnh (Quiescence Search) giải quyết chuỗi ăn quân và đe dọa chạm đích
    def qsearch(s, alpha, beta, ply, qd, h):
        nodes[0] += 1
        if (nodes[0] & 63) == 0 and (time.process_time() > deadline or time.process_time() - start_time > 0.84):
            abort[0] = True
        if abort[0]:
            return 0
        if wins_now(s):
            return WIN - ply
        o = 1 - s
        in_chk = wins_now(o)
        if in_chk:
            if qd >= 4:
                return -WIN + ply + 1
            all_m = gen(s, 0)
            def_sq = DEF_SQ[o]
            evasions = [m for m in all_m if m[3] in def_sq]
            if not evasions:
                return -WIN + ply + 1
            evasions.sort(reverse=True)
            best = -WIN + ply + 1
            for m in evasions:
                c = m[1]
                p = m[2]
                q = m[3]
                cap = m[4]
                board[p] = 0
                board[q] = c
                ps[c] = q
                h2 = h ^ Z[c][p] ^ Z[c][q] ^ ZS
                if cap:
                    ps[cap] = -1
                    cnt[o] -= 1
                    h2 ^= Z[cap][q]
                still_w = wins_now(o)
                if not still_w:
                    if cap and cnt[o] == 0:
                        val = WIN - ply
                    else:
                        val = -qsearch(o, -beta, -alpha, ply + 1, qd + 1, h2)
                else:
                    val = -WIN + ply + 1
                if cap:
                    cnt[o] += 1
                    ps[cap] = q
                board[q] = cap
                board[p] = c
                ps[c] = p
                if abort[0]:
                    return 0
                if val > best:
                    best = val
                    if val > alpha:
                        alpha = val
                        if alpha >= beta:
                            break
            return best
 
        stand = evaluate(s, h)
        if stand >= beta:
            return stand
        if stand > alpha:
            alpha = stand
        if qd >= 6:
            return stand
 
        moves = gen(s, 2 if qd == 0 else 1)
        if not moves:
            return stand
        moves.sort(reverse=True)
        for m in moves:
            c = m[1]
            p = m[2]
            q = m[3]
            cap = m[4]
            board[p] = 0
            board[q] = c
            ps[c] = q
            h2 = h ^ Z[c][p] ^ Z[c][q] ^ ZS
            if cap:
                ps[cap] = -1
                cnt[o] -= 1
                h2 ^= Z[cap][q]
            if cap and cnt[o] == 0:
                val = WIN - ply
            else:
                val = -qsearch(o, -beta, -alpha, ply + 1, qd + 1, h2)
            if cap:
                cnt[o] += 1
                ps[cap] = q
            board[q] = cap
            board[p] = c
            ps[c] = p
            if abort[0]:
                return 0
            if val > stand:
                stand = val
                if val > alpha:
                    alpha = val
                    if alpha >= beta:
                        break
        return stand
 
    # Alpha-Beta PVS kết hợp bảng băm TT, Killer Heuristic và Late Move Reduction
    def search(s, depth, alpha, beta, ply, h):
        nodes[0] += 1
        if (nodes[0] & 63) == 0 and (time.process_time() > deadline or time.process_time() - start_time > 0.84):
            abort[0] = True
        if abort[0]:
            return 0
        if wins_now(s):
            return WIN - ply
        o = 1 - s
        in_chk = wins_now(o)
        if depth <= 0:
            if in_chk and ply < PLYLIM:
                depth = 1
            else:
                return qsearch(s, alpha, beta, ply, 0, h)
        a0 = alpha
        ttm = None
        e = tt.get(h)
        if e is not None:
            ed, ef, ev, em = e
            if ev > WIN - 1000:
                ev -= ply
            elif ev < -WIN + 1000:
                ev += ply
            ttm = em
            if ed >= depth:
                if ef == 0:
                    return ev
                if ef == 1:
                    if ev >= beta:
                        return ev
                elif ev <= alpha:
                    return ev
 
        moves = gen(s, 0)
        if not moves:
            return -WIN + ply + 1 if in_chk else evaluate(s, h)
        if in_chk:
            def_sq = DEF_SQ[o]
            moves = [m for m in moves if m[3] in def_sq]
            if not moves:
                return -WIN + ply + 1
 
        moves.sort(reverse=True)
        km = KILLER[ply] if ply < 64 else None
        if km is not None or ttm is not None:
            for i in range(len(moves)):
                m = moves[i]
                if ttm is not None and m[1] == ttm[0] and m[2] == ttm[1] and m[3] == ttm[2]:
                    if i:
                        moves.insert(0, moves.pop(i))
                    break
            if km is not None:
                for i in range(1, len(moves)):
                    m = moves[i]
                    if m[1] == km[0] and m[2] == km[1] and m[3] == km[2]:
                        moves.insert(1 if ttm is not None else 0, moves.pop(i))
                        break
 
        best = -WIN + ply + 1 if in_chk else -INF
        bm = None
        idx = 0
        nb_set = DEF_SQ[s]
        for m in moves:
            c = m[1]
            p = m[2]
            q = m[3]
            cap = m[4]
            board[p] = 0
            board[q] = c
            ps[c] = q
            h2 = h ^ Z[c][p] ^ Z[c][q] ^ ZS
            if cap:
                ps[cap] = -1
                cnt[o] -= 1
                h2 ^= Z[cap][q]
 
            if in_chk and wins_now(o):
                val = -WIN + ply + 1
            elif cap and cnt[o] == 0:
                val = WIN - ply
            else:
                if idx == 0:
                    val = -search(o, depth - 1, -beta, -alpha, ply + 1, h2)
                else:
                    red = 1 if (idx >= 4 and depth >= 3 and not cap and not in_chk and q not in nb_set) else 0
                    if idx >= 12 and depth >= 5 and red:
                        red = 2
                    val = -search(o, depth - 1 - red, -alpha - 1, -alpha, ply + 1, h2)
                    if val > alpha and not abort[0]:
                        val = -search(o, depth - 1, -beta, -alpha, ply + 1, h2)
 
            if cap:
                cnt[o] += 1
                ps[cap] = q
            board[q] = cap
            board[p] = c
            ps[c] = p
            if abort[0]:
                return 0
            if val > best:
                best = val
                bm = (c, p, q)
                if val > alpha:
                    alpha = val
                    if alpha >= beta:
                        if not cap:
                            HIST[p * SIZE + q] += depth * depth
                            if ply < 64:
                                KILLER[ply] = bm
                        break
            idx += 1
 
        sv = best
        if sv > WIN - 1000:
            sv += ply
        elif sv < -WIN + 1000:
            sv -= ply
        if best <= a0:
            fl = 2
        elif best >= beta:
            fl = 1
        else:
            fl = 0
        tt[h] = (depth, fl, sv, bm)
        return best
 
    # Chọn nước đi tối ưu bằng Iterative Deepening tại nút gốc
    def decide():
        root = gen(0, 0)
        if not root:
            return (0, 0)
 
        # Kiểm tra nước thắng ngay lập tức
        gset0 = GOAL[0]
        for m in root:
            c0 = m[1]
            p0 = m[2]
            q0 = m[3]
            cap0 = m[4]
            if cap0 and cnt[1] == 1:
                return (pid_of[c0], DIRN[q0 - p0])
            if q0 in gset0:
                if goal_mode == 0:
                    return (pid_of[c0], DIRN[q0 - p0])
                elif all((g == q0 or (ALLY_LO <= board[g] <= ALLY_HI and g != p0)) for g in gset0):
                    return (pid_of[c0], DIRN[q0 - p0])
 
        # Nếu địch đang dọa chạm đích ở lượt kế, lọc các nước phòng thủ bắt buộc
        if wins_now(1):
            safe = []
            for m in root:
                c0 = m[1]
                p0 = m[2]
                q0 = m[3]
                cap0 = m[4]
                board[p0] = 0
                board[q0] = c0
                ps[c0] = q0
                if cap0:
                    ps[cap0] = -1
                    cnt[1] -= 1
                if not wins_now(1):
                    safe.append(m)
                if cap0:
                    cnt[1] += 1
                    ps[cap0] = q0
                board[q0] = cap0
                board[p0] = c0
                ps[c0] = p0
            if safe:
                root = safe
 
        if len(root) == 1:
            return (pid_of[root[0][1]], DIRN[root[0][3] - root[0][2]])
 
        h0 = 0
        for c in range(1, OPP_HI + 1):
            if ps[c] >= 0:
                h0 ^= Z[c][ps[c]]
 
        seen = memo.setdefault('seen', {})
        cfg = memo.setdefault('cfg', {})
        pen_unit = 130 + 2 * min(turn, 200)
        cfg_unit = (45 + 4 * stag) * (0.25 if behind else 1.0)
        hunt_range = max(10, (W + H) // 2 + 2)
 
        order = []
        for m in root:
            c = m[1]
            p = m[2]
            q = m[3]
            cap = m[4]
            t = tp[c]
            pred_t = PRED[t]
            prey_t = PREY[t]
            role = roles.get(c, 'HUNTER')
            tb = m[0] * 10
            if cap:
                tb += 5000
 
            min_pred_p = 999
            min_pred_q = 999
            min_prey_q = 999
            row_p = DT[p]
            row_q = DT[q]
            for ec in range(OPP_LO, OPP_HI + 1):
                ep = ps[ec]
                if ep >= 0 and ec != cap:
                    et = tp[ec]
                    if et == pred_t:
                        if row_p[ep] < min_pred_p:
                            min_pred_p = row_p[ep]
                        if row_q[ep] < min_pred_q:
                            min_pred_q = row_q[ep]
                    elif et == prey_t:
                        if row_q[ep] < min_prey_q:
                            min_prey_q = row_q[ep]
 
            if min_pred_p == 1 and min_pred_q > 1:
                tb += 350
            if min_pred_q == 1:
                tb -= 600
 
            if role == 'RUNNER':
                tb += (D0[p] - D0[q]) * 45 + (ADV0[q] - ADV0[p]) * 20
            elif role == 'DEFENDER':
                tb += (D1[p] - D1[q]) * 35
            else:
                tb += max(0, hunt_range - min_prey_q) * 25
            order.append((tb, m))
        order.sort(reverse=True)
        order = [x[1] for x in order]
 
        # Phạt nhẹ các thế cờ lặp lại để tránh đi quẩn quanh
        pens = {}
        for m in order:
            c = m[1]
            p = m[2]
            q = m[3]
            cap = m[4]
            board[p] = 0
            board[q] = c
            ps[c] = q
            if cap:
                ps[cap] = -1
            key = tuple(ps)
            ck = tuple(ps[ALLY_LO:ALLY_HI + 1])
            if cap:
                ps[cap] = q
            board[q] = cap
            board[p] = c
            ps[c] = p
            pens[(c, p, q)] = min(750, int(pen_unit * seen.get(key, 0) + cfg_unit * cfg.get(ck, 0)))
 
        best_move = order[0]
        for depth in range(1, 42):
            best_adj = -INF
            cur_best = None
            done = 0
            vals = []
            for m in order:
                if time.process_time() - start_time > 0.84:
                    abort[0] = True
                    break
                c = m[1]
                p = m[2]
                q = m[3]
                cap = m[4]
                pen = pens[(c, p, q)]
                araw = best_adj + pen if best_adj > -INF else -INF
                board[p] = 0
                board[q] = c
                ps[c] = q
                h2 = h0 ^ Z[c][p] ^ Z[c][q] ^ ZS
                if cap:
                    ps[cap] = -1
                    cnt[1] -= 1
                    h2 ^= Z[cap][q]
                val = -search(1, depth - 1, -INF, -araw if araw > -INF else INF, 1, h2)
                if cap:
                    cnt[1] += 1
                    ps[cap] = q
                board[q] = cap
                board[p] = c
                ps[c] = p
                if abort[0]:
                    break
                done += 1
                adj = val - pen if val < WIN - 1000 else val
                vals.append((adj, m))
                if adj > best_adj:
                    best_adj = adj
                    cur_best = m
 
            if cur_best is not None and (not abort[0] or done >= 1):
                best_move = cur_best
                if not abort[0]:
                    vals.sort(key=lambda z: z[0], reverse=True)
                    order = [t_item[1] for t_item in vals]
                else:
                    order.remove(cur_best)
                    order.insert(0, cur_best)
            if abort[0]:
                break
            if best_adj > WIN - 1000 or best_adj < -WIN + 1000:
                break
            if time.process_time() > deadline or time.process_time() - start_time > 0.84:
                break
 
        c = best_move[1]
        p = best_move[2]
        q = best_move[3]
        cap = best_move[4]
        board[p] = 0
        board[q] = c
        ps[c] = q
        if cap:
            ps[cap] = -1
        key = tuple(ps)
        seen[key] = seen.get(key, 0) + 1
        ck = tuple(ps[ALLY_LO:ALLY_HI + 1])
        cfg[ck] = cfg.get(ck, 0) + 1
        return (pid_of[c], DIRN[q - p])
 
    try:
        res = decide()
    except Exception:
        # Nhánh dự phòng an toàn nếu gặp sự cố bất ngờ
        res = (0, 0)
        bestk = -INF
        dirs = ((-1, 1), (0, 1), (1, 1), (1, 0), (1, -1), (0, -1), (-1, -1), (-1, 0))
        for p in ally_list:
            for d in range(1, 9):
                dx, dy = dirs[d - 1]
                nx = p.x + dx
                ny = p.y + dy
                if nx < 1 or nx > W or ny < 1 or ny > H or (nx, ny) in walls_xy:
                    continue
                ok = True
                for a in ally_list:
                    if a.x == nx and a.y == ny:
                        ok = False
                for e in opp_list:
                    if e.x == nx and e.y == ny:
                        if not ((p.Type == 'R' and e.Type == 'S') or (p.Type == 'S' and e.Type == 'P') or (p.Type == 'P' and e.Type == 'R')):
                            ok = False
                if ok and ADV0[ny * STRIDE + nx] > bestk:
                    bestk = ADV0[ny * STRIDE + nx]
                    res = (p.pid, d)
 
    memo['used'] = used + (time.process_time() - start_time)
    return res

import math
import random
import time


def play(gameState, memo={}):
    start_time = time.process_time()
    turn = gameState.currentTurn
    maxT = gameState.maxTurn

    INF = 10 ** 9
    WIN = 100000
    PLYLIM = 44

    W = gameState.width
    H = gameState.height
    goal_mode = gameState.goalMode

    walls_xy = frozenset((w.x, w.y) for w in gameState.walls)
    goals0_xy = frozenset((g.x, g.y) for g in gameState.goals) or frozenset([(W, H)])
    goals1_xy = frozenset((g.x, g.y) for g in gameState.enemyGoals) or frozenset([(1, 1)])

    # Bảng khắc chế: 0=R (Búa), 1=P (Bao), 2=S (Kéo)
    BT = ((False, False, True), (True, False, False), (False, True, False))
    PRED = (1, 2, 0)
    PREY = (2, 0, 1)

    ELIM1 = 65
    ELIM2 = 0

    # Nếu trận trước bị reset giữa chừng thì xóa cache cũ
    if turn <= 1 and 'tabs' in memo:
        memo.clear()

    map_sig = (W, H, walls_xy, goals0_xy, goals1_xy, goal_mode)
    tabs = memo.get('tabs')

    # Khởi tạo bảng tra cứu bản đồ (chạy 1 lần ở lượt đầu hoặc khi đổi map)
    if tabs is None or memo.get('map_sig') != map_sig:
        memo['map_sig'] = map_sig
        memo['tt'] = {}
        memo['hist'] = None
        memo['seen'] = {}
        memo['cfg'] = {}

        STRIDE = W + 2
        SIZE = (W + 2) * (H + 2)
        OFFS = (STRIDE + 1, STRIDE, 1, STRIDE - 1, -(STRIDE - 1), -1, -STRIDE, -(STRIDE + 1))
        DIRN = {
            STRIDE - 1: 1,
            STRIDE: 2,
            STRIDE + 1: 3,
            1: 4,
            -(STRIDE - 1): 5,
            -STRIDE: 6,
            -(STRIDE + 1): 7,
            -1: 8,
        }

        wall_set = frozenset(
            y * STRIDE + x for (x, y) in walls_xy if 1 <= x <= W and 1 <= y <= H
        )
        cells = [
            y * STRIDE + x
            for y in range(1, H + 1)
            for x in range(1, W + 1)
            if (y * STRIDE + x) not in wall_set
        ]
        valid_set = set(cells)

        GOAL0 = frozenset(y * STRIDE + x for (x, y) in goals0_xy if (y * STRIDE + x) in valid_set)
        GOAL1 = frozenset(y * STRIDE + x for (x, y) in goals1_xy if (y * STRIDE + x) in valid_set)
        GOAL = (GOAL0, GOAL1)

        nb0_set = {g + off for g in GOAL0 for off in OFFS if (g + off) in valid_set and (g + off) not in GOAL0}
        nb1_set = {g + off for g in GOAL1 for off in OFFS if (g + off) in valid_set and (g + off) not in GOAL1}
        NB = (tuple(nb0_set), tuple(nb1_set))
        DEF_SQ = (GOAL0 | nb0_set, GOAL1 | nb1_set)

        MAX_D = W + H + 12

        # BFS tính khoảng cách ngắn nhất né tường tới tập ô đích của ta (D0) và địch (D1)
        D0 = [MAX_D] * SIZE
        bfs_q = list(GOAL0)
        for g in GOAL0:
            D0[g] = 0
        head = 0
        while head < len(bfs_q):
            u = bfs_q[head]
            head += 1
            du = D0[u] + 1
            for off in OFFS:
                v = u + off
                if v in valid_set and du < D0[v]:
                    D0[v] = du
                    bfs_q.append(v)

        D1 = [MAX_D] * SIZE
        bfs_q = list(GOAL1)
        for g in GOAL1:
            D1[g] = 0
        head = 0
        while head < len(bfs_q):
            u = bfs_q[head]
            head += 1
            du = D1[u] + 1
            for off in OFFS:
                v = u + off
                if v in valid_set and du < D1[v]:
                    D1[v] = du
                    bfs_q.append(v)

        # BFS tính khoảng cách thực tế giữa mọi cặp ô hợp lệ để săn đuổi vòng qua vật cản
        DT = [[MAX_D] * SIZE for _ in range(SIZE)]
        MDT = [[MAX_D * 2] * SIZE for _ in range(SIZE)]
        for src in cells:
            row = DT[src]
            mrow = MDT[src]
            row[src] = 0
            mrow[src] = 0
            bfs_q = [src]
            head = 0
            while head < len(bfs_q):
                u = bfs_q[head]
                head += 1
                du = row[u] + 1
                for off in OFFS:
                    v = u + off
                    if v in valid_set and du < row[v]:
                        row[v] = du
                        mrow[v] = du * 2
                        bfs_q.append(v)

        max_d0 = max((D0[a] for a in cells if D0[a] < MAX_D), default=1)
        max_d1 = max((D1[a] for a in cells if D1[a] < MAX_D), default=1)
        ADV0 = [0] * SIZE
        ADV1 = [0] * SIZE
        CENTER = [0] * SIZE
        mid_x = (W + 1) * 0.5
        mid_y = (H + 1) * 0.5
        for a in cells:
            x = a % STRIDE
            y = a // STRIDE
            ADV0[a] = max(0, max_d0 - D0[a])
            ADV1[a] = max(0, max_d1 - D1[a])
            cx = abs(x - mid_x) / max(1.0, W / 9.0)
            cy = abs(y - mid_y) / max(1.0, H / 9.0)
            CENTER[a] = max(0, int(4 - max(cx, cy)))

        # Bảng nước đi hợp lệ sắp xếp sẵn theo độ giảm khoảng cách tới đích
        MTAB0 = [() for _ in range(SIZE)]
        MTAB1 = [() for _ in range(SIZE)]
        for a in cells:
            m0 = []
            m1 = []
            for off in OFFS:
                q = a + off
                if q in valid_set:
                    h_ix = a * SIZE + q
                    fw0_b = (D0[a] - D0[q]) * 2
                    fw1_b = (D1[a] - D1[q]) * 2
                    m0.append((q, fw0_b, h_ix, 1 if q in nb0_set else 0))
                    m1.append((q, fw1_b, h_ix, 1 if q in nb1_set else 0))
            m0.sort(key=lambda item: item[1], reverse=True)
            m1.sort(key=lambda item: item[1], reverse=True)
            MTAB0[a] = tuple(m0)
            MTAB1[a] = tuple(m1)
        MTAB = (tuple(MTAB0), tuple(MTAB1))

        # Bảng điểm thưởng khoảng cách tới đích (tự co giãn theo kích thước map)
        RT = tuple([0] + [int(950 * (0.60 ** (d - 1))) for d in range(1, MAX_D + 1)])
        UNSTOP_RT = tuple([0] + [int(1750 * (0.64 ** (d - 1))) for d in range(1, MAX_D + 1)])

        rz = random.Random(20260402)
        Z = tuple(tuple(rz.getrandbits(52) for _ in range(SIZE)) for _ in range(48))
        ZS = rz.getrandbits(52)

        tabs = (
            STRIDE, SIZE, MAX_D, OFFS, DIRN, wall_set, GOAL, NB, DEF_SQ,
            DT, MDT, ADV0, ADV1, D0, D1, CENTER, MTAB, RT, UNSTOP_RT, Z, ZS
        )
        memo['tabs'] = tabs

    (
        STRIDE, SIZE, MAX_D, OFFS, DIRN, wall_set, GOAL, NB, DEF_SQ,
        DT, MDT, ADV0, ADV1, D0, D1, CENTER, MTAB, RT, UNSTOP_RT, Z, ZS
    ) = tabs

    # Quản lý ngân sách CPU: tổng 180s/ván, tối đa 1s/nước (chốt ở 0.82s để an toàn)
    used = memo.get('used', 0.0)
    left = max(1, (maxT - turn) // 2 + 1)
    if turn <= 1:
        budget = 0.82
    else:
        budget = min(0.82, max(0.04, (160.0 - used) / min(left, 150)))
    deadline = start_time + budget

    # Dựng bàn cờ 1 chiều: ngoài biên và ô tường đen mang giá trị -1
    board = [-1] * SIZE
    for yy in range(1, H + 1):
        base_y = yy * STRIDE
        for xx in range(1, W + 1):
            pos = base_y + xx
            if pos not in wall_set:
                board[pos] = 0

    # Đánh chỉ số động cho quân hai bên (hỗ trợ map có số quân khác 9)
    ally_list = list(gameState.ally)
    opp_list = list(gameState.opponent)
    na_init = len(ally_list)
    nb_init = len(opp_list)

    ALLY_LO = 1
    ALLY_HI = na_init
    OPP_LO = na_init + 2
    OPP_HI = OPP_LO + nb_init - 1
    LO = (ALLY_LO, OPP_LO)
    HI = (ALLY_HI, OPP_HI)

    slot_cap = max(48, OPP_HI + 2)
    if len(Z) < slot_cap:
        rz = random.Random(20260402)
        Z = tuple(tuple(rz.getrandbits(52) for _ in range(SIZE)) for _ in range(slot_cap))

    tp = [0] * slot_cap
    ps = [-1] * slot_cap
    pid_of = [0] * slot_cap
    TI = {'R': 0, 'P': 1, 'S': 2}
    cnt = [0, 0]

    for idx_a, p in enumerate(ally_list):
        c = ALLY_LO + idx_a
        pid_of[c] = p.pid
        tp[c] = TI[p.Type]
        pos = p.y * STRIDE + p.x
        ps[c] = pos
        board[pos] = c
        cnt[0] += 1

    for idx_b, p in enumerate(opp_list):
        c = OPP_LO + idx_b
        pid_of[c] = p.pid
        tp[c] = TI[p.Type]
        pos = p.y * STRIDE + p.x
        ps[c] = pos
        board[pos] = c
        cnt[1] += 1

    tt = memo.get('tt')
    if tt is None or len(tt) > 120000:
        tt = {}
        memo['tt'] = tt

    hist_len = SIZE * SIZE
    HIST = memo.get('hist')
    if HIST is None or len(HIST) != hist_len:
        HIST = [0] * hist_len
        memo['hist'] = HIST
    else:
        for i in range(hist_len):
            if HIST[i]:
                HIST[i] >>= 1

    KILLER = [None] * 64
    nodes = [0]
    abort = [False]

    total = cnt[0] + cnt[1]
    if memo.get('prev_total') != total:
        memo['prev_total'] = total
        memo['capt_turn'] = turn
        memo['cfg'] = {}
    stag = (turn - memo.get('capt_turn', turn)) // 2
    ahead = cnt[0] > cnt[1]
    behind = cnt[0] < cnt[1]

    K = 5 + min(8, turn // 25)
    if not behind and stag > 16:
        K += min(7, (stag - 16) // 6)
    if behind:
        HW = 3
    elif stag > 16:
        HW = 11
    else:
        HW = 6
    if ahead and stag > 26:
        DP = (0, 220, 120, 55)
    else:
        DP = (0, 330, 175, 80)

    # Phân vai chiến thuật: DEFENDER (chặn đích nhà), RUNNER (đột phá), HUNTER (săn quân)
    roles = {}
    e_types_alive = [0, 0, 0]
    e_near_home = []
    e_near_away = [0, 0, 0]
    for c in range(OPP_LO, OPP_HI + 1):
        q = ps[c]
        if q >= 0:
            u = tp[c]
            e_types_alive[u] += 1
            if D1[q] <= 4:
                e_near_home.append((D1[q], q, u))
            if D0[q] <= 2:
                e_near_away[u] += 1
    e_near_home.sort()

    def_assigned = set()
    for d_h, q_h, u_h in e_near_home:
        need_t = PRED[u_h]
        best_c = 0
        best_d = 999
        for c in range(ALLY_LO, ALLY_HI + 1):
            p = ps[c]
            if p >= 0 and tp[c] == need_t and c not in def_assigned:
                d_c = D1[p]
                if d_c < best_d:
                    best_d = d_c
                    best_c = c
        if best_c:
            roles[best_c] = 'DEFENDER'
            def_assigned.add(best_c)

    if not def_assigned:
        best_c = 0
        best_d = 999
        for c in range(ALLY_LO, ALLY_HI + 1):
            p = ps[c]
            if p >= 0 and e_types_alive[PREY[tp[c]]] > 0:
                d_c = D1[p]
                if d_c < best_d:
                    best_d = d_c
                    best_c = c
        if best_c and best_d <= 3:
            roles[best_c] = 'DEFENDER'
            def_assigned.add(best_c)

    runner_cands = []
    for c in range(ALLY_LO, ALLY_HI + 1):
        p = ps[c]
        if p >= 0 and c not in def_assigned:
            t = tp[c]
            imm = 1 if e_types_alive[PRED[t]] == 0 else 0
            brk = 1 if e_near_away[PREY[t]] > 0 else 0
            runner_cands.append((D0[p] - 2 * imm - brk, -ADV0[p], c))
    runner_cands.sort()
    num_runners = max(2, len(GOAL[0])) if goal_mode == 1 else 2
    for idx_r in range(min(num_runners, len(runner_cands))):
        roles[runner_cands[idx_r][2]] = 'RUNNER'
    for c in range(ALLY_LO, ALLY_HI + 1):
        if ps[c] >= 0 and c not in roles:
            roles[c] = 'HUNTER'

    # Kiểm tra phe s có thể chạm đích thắng ngay trong 1 nước đi không
    def wins_now(s):
        lo = LO[s]
        hi = HI[s]
        gset = GOAL[s]
        if goal_mode == 0:
            for g in gset:
                v0 = board[g]
                if lo <= v0 <= hi:
                    return True
                if v0 == 0:
                    for off in OFFS:
                        v = board[g + off]
                        if lo <= v <= hi:
                            return True
                else:
                    t = tp[v0]
                    for off in OFFS:
                        v = board[g + off]
                        if lo <= v <= hi and BT[tp[v]][t]:
                            return True
            return False
        else:
            missing = [g for g in gset if not (lo <= board[g] <= hi)]
            if not missing:
                return True
            if len(missing) > 1:
                return False
            g = missing[0]
            v0 = board[g]
            for off in OFFS:
                src = g + off
                v = board[src]
                if lo <= v <= hi and src not in gset:
                    if v0 == 0 or BT[tp[v]][tp[v0]]:
                        return True
            return False

    # Sinh nước đi hợp lệ: mode=0 (tất cả), mode=1 (chỉ ăn quân), mode=2 (ăn quân + áp sát đích)
    def gen(s, mode):
        res = []
        lo = LO[s]
        hi = HI[s]
        elo = LO[1 - s]
        ehi = HI[1 - s]
        mtab = MTAB[s]
        gset = GOAL[s]
        egset = GOAL[1 - s]
        for c in range(lo, hi + 1):
            p = ps[c]
            if p < 0:
                continue
            t = tp[c]
            bt = BT[t]
            on_goal = (goal_mode == 1 and p in gset)
            for q, fw_b, h_ix, is_nb in mtab[p]:
                v = board[q]
                if v == 0:
                    if mode == 0:
                        h = HIST[h_ix]
                        pen_leave = -35 if (on_goal and q not in gset) else 0
                        res.append((fw_b + pen_leave + (h if h < 72 else 72), c, p, q, 0))
                    elif mode == 2 and is_nb:
                        for off in OFFS:
                            gq = q + off
                            if gq in gset:
                                vg = board[gq]
                                if vg == 0 or (elo <= vg <= ehi and bt[tp[vg]]):
                                    res.append((820 + fw_b, c, p, q, 0))
                                    break
                elif elo <= v <= ehi:
                    if bt[tp[v]]:
                        bonus = 90 if (q in egset or D1[q] <= 2) else 0
                        res.append((1050 + fw_b + bonus, c, p, q, v))
        return res

    # Hàm lượng giá thế cờ tĩnh từ góc nhìn phe s
    def evaluate_raw(s):
        A_p = []
        A_t = []
        B_p = []
        B_t = []
        cA0 = cA1 = cA2 = 0
        cB0 = cB1 = cB2 = 0
        for c in range(ALLY_LO, ALLY_HI + 1):
            p = ps[c]
            if p >= 0:
                t = tp[c]
                A_p.append(p)
                A_t.append(t)
                if t == 0:
                    cA0 += 1
                elif t == 1:
                    cA1 += 1
                else:
                    cA2 += 1
        for c in range(OPP_LO, OPP_HI + 1):
            p = ps[c]
            if p >= 0:
                t = tp[c]
                B_p.append(p)
                B_t.append(t)
                if t == 0:
                    cB0 += 1
                elif t == 1:
                    cB1 += 1
                else:
                    cB2 += 1

        na = len(A_p)
        nb = len(B_p)
        if na == 0:
            return -WIN + 100 if s == 0 else WIN - 100
        if nb == 0:
            return WIN - 100 if s == 0 else -WIN + 100

        cA = (cA0, cA1, cA2)
        cB = (cB0, cB1, cB2)

        # Hơn quân và lợi thế tuyệt đối khi địch mất sạch loại quân khắc chế
        v = 1050 * (na - nb) + 3200 * (na - nb) // (na + nb)

        for t in (0, 1, 2):
            pred = PRED[t]
            prey = PREY[t]
            ca = cA[t]
            cb = cB[t]
            if ca:
                if cB[pred] == 0:
                    v += ca * (500 if cB[t] == 0 else 360)
                elif cB[pred] == 1 and cA[prey] > 0:
                    v += ca * ELIM1
                elif ELIM2 and cB[pred] == 2 and cA[prey] >= 2:
                    v += ca * ELIM2
                v += ca * (80 * cB[prey] if cB[prey] else -50)
            if cb:
                if cA[pred] == 0:
                    v -= cb * (500 if cA[t] == 0 else 360)
                elif cA[pred] == 1 and cB[prey] > 0:
                    v -= cb * ELIM1
                elif ELIM2 and cA[pred] == 2 and cB[prey] >= 2:
                    v -= cb * ELIM2
                v -= cb * (80 * cA[prey] if cA[prey] else -50)

        # Điểm chiếm đóng đồng thời khi bản đồ bật goalMode == 1
        if goal_mode == 1:
            occ0 = 0
            for g in GOAL[0]:
                if ALLY_LO <= board[g] <= ALLY_HI:
                    occ0 += 1
            occ1 = 0
            for g in GOAL[1]:
                if OPP_LO <= board[g] <= OPP_HI:
                    occ1 += 1
            if occ0 == len(GOAL[0]):
                return WIN - 50 if s == 0 else -WIN + 50
            if occ1 == len(GOAL[1]):
                return -WIN + 50 if s == 0 else WIN - 50
            v += 900 * (occ0 - occ1)

        ha = [MAX_D] * na
        pa = [MAX_D] * na
        ma = [MAX_D * 2] * na
        inta = [999] * na
        hb = [MAX_D] * nb
        pb = [MAX_D] * nb
        mb = [MAX_D * 2] * nb
        intb = [999] * nb

        for i in range(na):
            p = A_p[i]
            t = A_t[i]
            d0p = D0[p]
            row = DT[p]
            mrow = MDT[p]
            bt = BT[t]
            for j in range(nb):
                q = B_p[j]
                u = B_t[j]
                d0q = D0[q]
                d1q = D1[q]
                if u == t:
                    if d0q <= d0p and d0q < inta[i]:
                        inta[i] = d0q
                    if D1[p] <= d1q and D1[p] < intb[j]:
                        intb[j] = D1[p]
                    continue
                d = row[q]
                md = mrow[q]
                if bt[u]:
                    if d < pa[i]:
                        pa[i] = d
                    if md < ma[i]:
                        ma[i] = md
                    if d < hb[j]:
                        hb[j] = d
                    if D1[p] <= d1q + 1 and D1[p] < intb[j]:
                        intb[j] = D1[p]
                else:
                    if d < ha[i]:
                        ha[i] = d
                    if d < pb[j]:
                        pb[j] = d
                    if md < mb[j]:
                        mb[j] = md
                    if d0q <= d0p + 1 and d0q < inta[i]:
                        inta[i] = d0q

        adv0 = 0
        thr0 = 0
        dg0 = 0
        hunt0 = 0
        hang0 = 0
        away_guard_b = [0, 0, 0]
        for j in range(nb):
            if D0[B_p[j]] <= 2:
                away_guard_b[B_t[j]] = 1

        max_unstop = len(UNSTOP_RT) - 1
        max_rt = len(RT) - 1
        hunt_range = max(10, (W + H) // 2 + 2)

        for i in range(na):
            p = A_p[i]
            t = A_t[i]
            ap = ADV0[p]
            if cB[PRED[t]] == 0:
                ap += ap >> 1
            if away_guard_b[PREY[t]]:
                ap += max(0, hunt_range - D0[p]) * 3
            adv0 += ap

            mp = pa[i]
            if mp < MAX_D:
                hunt0 += HW * max(0, hunt_range - mp) + max(0, hunt_range * 2 - ma[i])
            else:
                hunt0 += CENTER[p]

            mh = ha[i]
            if mh <= 3:
                pt = PREY[t]
                row = DT[p]
                prot = False
                for k in range(na):
                    if A_t[k] == pt and row[A_p[k]] <= 1:
                        prot = True
                        break
                if not prot:
                    if mh <= 1:
                        dg0 += DP[1] + (160 if s == 1 else 0)
                        hang0 += 1
                    else:
                        dg0 += DP[mh]
                elif mh <= 2:
                    hunt0 += 25

            d = D0[p]
            if inta[i] == 999:
                if d <= max_unstop:
                    tv = UNSTOP_RT[d]
                    if tv > thr0:
                        thr0 = tv
            elif d <= max_rt:
                tv = RT[d]
                if mh <= 2 or inta[i] <= d:
                    tv //= 3
                if tv > thr0:
                    thr0 = tv

        if hang0 >= 2:
            dg0 += 380 * (hang0 - 1)

        adv1 = 0
        thr1 = 0
        dg1 = 0
        hunt1 = 0
        hang1 = 0
        home_guard_a = [0, 0, 0]
        for i in range(na):
            if D1[A_p[i]] <= 2:
                home_guard_a[A_t[i]] = 1

        for j in range(nb):
            q = B_p[j]
            u = B_t[j]
            aq = ADV1[q]
            if cA[PRED[u]] == 0:
                aq += aq >> 1
            if home_guard_a[PREY[u]]:
                aq += max(0, hunt_range - D1[q]) * 3
            adv1 += aq

            mp = pb[j]
            if mp < MAX_D:
                hunt1 += HW * max(0, hunt_range - mp) + max(0, hunt_range * 2 - mb[j])
            else:
                hunt1 += CENTER[q]

            mh = hb[j]
            if mh <= 3:
                pt = PREY[u]
                row = DT[q]
                prot = False
                for k in range(nb):
                    if B_t[k] == pt and row[B_p[k]] <= 1:
                        prot = True
                        break
                if not prot:
                    if mh <= 1:
                        dg1 += DP[1] + (160 if s == 0 else 0)
                        hang1 += 1
                    else:
                        dg1 += DP[mh]
                elif mh <= 2:
                    hunt1 += 25

            d = D1[q]
            if intb[j] == 999:
                if d <= max_unstop:
                    tv = UNSTOP_RT[d]
                    if tv > thr1:
                        thr1 = tv
            elif d <= max_rt:
                tv = RT[d]
                if mh <= 2 or intb[j] <= d:
                    tv //= 3
                if tv > thr1:
                    thr1 = tv

        if hang1 >= 2:
            dg1 += 380 * (hang1 - 1)

        v += K * (adv0 - adv1) + (thr0 - thr1) - (dg0 - dg1) + (hunt0 - hunt1)

        # Phòng thủ vùng cờ nhà
        has_useful_def = False
        for i in range(na):
            if D1[A_p[i]] <= 2 and cB[PREY[A_t[i]]] > 0:
                has_useful_def = True
                break

        near_cnt = 0
        for j in range(nb):
            d = D1[B_p[j]]
            if d <= 2:
                v -= 520 * (3 - d)
                near_cnt += 1
                if intb[j] == 999:
                    v -= 450 * (3 - d)
            elif d == 3:
                v -= 110
                if intb[j] == 999:
                    v -= 220
            elif d == 4:
                v -= 35

        if near_cnt >= 2:
            v -= 420 * (near_cnt - 1)
        if near_cnt >= 1 and not has_useful_def:
            v -= 420 * near_cnt
        if has_useful_def:
            v += 135 + 105 * near_cnt

        return v if s == 0 else -v

    evc = {}

    def evaluate(s, h):
        r = evc.get(h)
        if r is None:
            r = evaluate_raw(s)
            evc[h] = r
        return r

    # Tìm kiếm tĩnh (Quiescence Search) giải quyết chuỗi ăn quân và đe dọa chạm đích
    def qsearch(s, alpha, beta, ply, qd, h):
        nodes[0] += 1
        if (nodes[0] & 63) == 0 and (time.process_time() > deadline or time.process_time() - start_time > 0.84):
            abort[0] = True
        if abort[0]:
            return 0
        if wins_now(s):
            return WIN - ply
        o = 1 - s
        in_chk = wins_now(o)
        if in_chk:
            if qd >= 4:
                return -WIN + ply + 1
            all_m = gen(s, 0)
            def_sq = DEF_SQ[o]
            evasions = [m for m in all_m if m[3] in def_sq]
            if not evasions:
                return -WIN + ply + 1
            evasions.sort(reverse=True)
            best = -WIN + ply + 1
            for m in evasions:
                c = m[1]
                p = m[2]
                q = m[3]
                cap = m[4]
                board[p] = 0
                board[q] = c
                ps[c] = q
                h2 = h ^ Z[c][p] ^ Z[c][q] ^ ZS
                if cap:
                    ps[cap] = -1
                    cnt[o] -= 1
                    h2 ^= Z[cap][q]
                still_w = wins_now(o)
                if not still_w:
                    if cap and cnt[o] == 0:
                        val = WIN - ply
                    else:
                        val = -qsearch(o, -beta, -alpha, ply + 1, qd + 1, h2)
                else:
                    val = -WIN + ply + 1
                if cap:
                    cnt[o] += 1
                    ps[cap] = q
                board[q] = cap
                board[p] = c
                ps[c] = p
                if abort[0]:
                    return 0
                if val > best:
                    best = val
                    if val > alpha:
                        alpha = val
                        if alpha >= beta:
                            break
            return best

        stand = evaluate(s, h)
        if stand >= beta:
            return stand
        if stand > alpha:
            alpha = stand
        if qd >= 6:
            return stand

        moves = gen(s, 2 if qd == 0 else 1)
        if not moves:
            return stand
        moves.sort(reverse=True)
        for m in moves:
            c = m[1]
            p = m[2]
            q = m[3]
            cap = m[4]
            board[p] = 0
            board[q] = c
            ps[c] = q
            h2 = h ^ Z[c][p] ^ Z[c][q] ^ ZS
            if cap:
                ps[cap] = -1
                cnt[o] -= 1
                h2 ^= Z[cap][q]
            if cap and cnt[o] == 0:
                val = WIN - ply
            else:
                val = -qsearch(o, -beta, -alpha, ply + 1, qd + 1, h2)
            if cap:
                cnt[o] += 1
                ps[cap] = q
            board[q] = cap
            board[p] = c
            ps[c] = p
            if abort[0]:
                return 0
            if val > stand:
                stand = val
                if val > alpha:
                    alpha = val
                    if alpha >= beta:
                        break
        return stand

    # Alpha-Beta PVS kết hợp bảng băm TT, Killer Heuristic và Late Move Reduction
    def search(s, depth, alpha, beta, ply, h):
        nodes[0] += 1
        if (nodes[0] & 63) == 0 and (time.process_time() > deadline or time.process_time() - start_time > 0.84):
            abort[0] = True
        if abort[0]:
            return 0
        if wins_now(s):
            return WIN - ply
        o = 1 - s
        in_chk = wins_now(o)
        if depth <= 0:
            if in_chk and ply < PLYLIM:
                depth = 1
            else:
                return qsearch(s, alpha, beta, ply, 0, h)
        a0 = alpha
        ttm = None
        e = tt.get(h)
        if e is not None:
            ed, ef, ev, em = e
            if ev > WIN - 1000:
                ev -= ply
            elif ev < -WIN + 1000:
                ev += ply
            ttm = em
            if ed >= depth:
                if ef == 0:
                    return ev
                if ef == 1:
                    if ev >= beta:
                        return ev
                elif ev <= alpha:
                    return ev

        moves = gen(s, 0)
        if not moves:
            return -WIN + ply + 1 if in_chk else evaluate(s, h)
        if in_chk:
            def_sq = DEF_SQ[o]
            moves = [m for m in moves if m[3] in def_sq]
            if not moves:
                return -WIN + ply + 1

        moves.sort(reverse=True)
        km = KILLER[ply] if ply < 64 else None
        if km is not None or ttm is not None:
            for i in range(len(moves)):
                m = moves[i]
                if ttm is not None and m[1] == ttm[0] and m[2] == ttm[1] and m[3] == ttm[2]:
                    if i:
                        moves.insert(0, moves.pop(i))
                    break
            if km is not None:
                for i in range(1, len(moves)):
                    m = moves[i]
                    if m[1] == km[0] and m[2] == km[1] and m[3] == km[2]:
                        moves.insert(1 if ttm is not None else 0, moves.pop(i))
                        break

        best = -WIN + ply + 1 if in_chk else -INF
        bm = None
        idx = 0
        nb_set = DEF_SQ[s]
        for m in moves:
            c = m[1]
            p = m[2]
            q = m[3]
            cap = m[4]
            board[p] = 0
            board[q] = c
            ps[c] = q
            h2 = h ^ Z[c][p] ^ Z[c][q] ^ ZS
            if cap:
                ps[cap] = -1
                cnt[o] -= 1
                h2 ^= Z[cap][q]

            if in_chk and wins_now(o):
                val = -WIN + ply + 1
            elif cap and cnt[o] == 0:
                val = WIN - ply
            else:
                if idx == 0:
                    val = -search(o, depth - 1, -beta, -alpha, ply + 1, h2)
                else:
                    red = 1 if (idx >= 4 and depth >= 3 and not cap and not in_chk and q not in nb_set) else 0
                    if idx >= 12 and depth >= 5 and red:
                        red = 2
                    val = -search(o, depth - 1 - red, -alpha - 1, -alpha, ply + 1, h2)
                    if val > alpha and not abort[0]:
                        val = -search(o, depth - 1, -beta, -alpha, ply + 1, h2)

            if cap:
                cnt[o] += 1
                ps[cap] = q
            board[q] = cap
            board[p] = c
            ps[c] = p
            if abort[0]:
                return 0
            if val > best:
                best = val
                bm = (c, p, q)
                if val > alpha:
                    alpha = val
                    if alpha >= beta:
                        if not cap:
                            HIST[p * SIZE + q] += depth * depth
                            if ply < 64:
                                KILLER[ply] = bm
                        break
            idx += 1

        sv = best
        if sv > WIN - 1000:
            sv += ply
        elif sv < -WIN + 1000:
            sv -= ply
        if best <= a0:
            fl = 2
        elif best >= beta:
            fl = 1
        else:
            fl = 0
        tt[h] = (depth, fl, sv, bm)
        return best

    # Chọn nước đi tối ưu bằng Iterative Deepening tại nút gốc
    def decide():
        root = gen(0, 0)
        if not root:
            return (0, 0)

        # Kiểm tra nước thắng ngay lập tức
        gset0 = GOAL[0]
        for m in root:
            c0 = m[1]
            p0 = m[2]
            q0 = m[3]
            cap0 = m[4]
            if cap0 and cnt[1] == 1:
                return (pid_of[c0], DIRN[q0 - p0])
            if q0 in gset0:
                if goal_mode == 0:
                    return (pid_of[c0], DIRN[q0 - p0])
                elif all((g == q0 or (ALLY_LO <= board[g] <= ALLY_HI and g != p0)) for g in gset0):
                    return (pid_of[c0], DIRN[q0 - p0])

        # Nếu địch đang dọa chạm đích ở lượt kế, lọc các nước phòng thủ bắt buộc
        if wins_now(1):
            safe = []
            for m in root:
                c0 = m[1]
                p0 = m[2]
                q0 = m[3]
                cap0 = m[4]
                board[p0] = 0
                board[q0] = c0
                ps[c0] = q0
                if cap0:
                    ps[cap0] = -1
                    cnt[1] -= 1
                if not wins_now(1):
                    safe.append(m)
                if cap0:
                    cnt[1] += 1
                    ps[cap0] = q0
                board[q0] = cap0
                board[p0] = c0
                ps[c0] = p0
            if safe:
                root = safe

        if len(root) == 1:
            return (pid_of[root[0][1]], DIRN[root[0][3] - root[0][2]])

        h0 = 0
        for c in range(1, OPP_HI + 1):
            if ps[c] >= 0:
                h0 ^= Z[c][ps[c]]

        seen = memo.setdefault('seen', {})
        cfg = memo.setdefault('cfg', {})
        pen_unit = 130 + 2 * min(turn, 200)
        cfg_unit = (45 + 4 * stag) * (0.25 if behind else 1.0)
        hunt_range = max(10, (W + H) // 2 + 2)

        order = []
        for m in root:
            c = m[1]
            p = m[2]
            q = m[3]
            cap = m[4]
            t = tp[c]
            pred_t = PRED[t]
            prey_t = PREY[t]
            role = roles.get(c, 'HUNTER')
            tb = m[0] * 10
            if cap:
                tb += 5000

            min_pred_p = 999
            min_pred_q = 999
            min_prey_q = 999
            row_p = DT[p]
            row_q = DT[q]
            for ec in range(OPP_LO, OPP_HI + 1):
                ep = ps[ec]
                if ep >= 0 and ec != cap:
                    et = tp[ec]
                    if et == pred_t:
                        if row_p[ep] < min_pred_p:
                            min_pred_p = row_p[ep]
                        if row_q[ep] < min_pred_q:
                            min_pred_q = row_q[ep]
                    elif et == prey_t:
                        if row_q[ep] < min_prey_q:
                            min_prey_q = row_q[ep]

            if min_pred_p == 1 and min_pred_q > 1:
                tb += 350
            if min_pred_q == 1:
                tb -= 600

            if role == 'RUNNER':
                tb += (D0[p] - D0[q]) * 45 + (ADV0[q] - ADV0[p]) * 20
            elif role == 'DEFENDER':
                tb += (D1[p] - D1[q]) * 35
            else:
                tb += max(0, hunt_range - min_prey_q) * 25
            order.append((tb, m))
        order.sort(reverse=True)
        order = [x[1] for x in order]

        # Phạt nhẹ các thế cờ lặp lại để tránh đi quẩn quanh
        pens = {}
        for m in order:
            c = m[1]
            p = m[2]
            q = m[3]
            cap = m[4]
            board[p] = 0
            board[q] = c
            ps[c] = q
            if cap:
                ps[cap] = -1
            key = tuple(ps)
            ck = tuple(ps[ALLY_LO:ALLY_HI + 1])
            if cap:
                ps[cap] = q
            board[q] = cap
            board[p] = c
            ps[c] = p
            pens[(c, p, q)] = min(750, int(pen_unit * seen.get(key, 0) + cfg_unit * cfg.get(ck, 0)))

        best_move = order[0]
        for depth in range(1, 42):
            best_adj = -INF
            cur_best = None
            done = 0
            vals = []
            for m in order:
                if time.process_time() - start_time > 0.84:
                    abort[0] = True
                    break
                c = m[1]
                p = m[2]
                q = m[3]
                cap = m[4]
                pen = pens[(c, p, q)]
                araw = best_adj + pen if best_adj > -INF else -INF
                board[p] = 0
                board[q] = c
                ps[c] = q
                h2 = h0 ^ Z[c][p] ^ Z[c][q] ^ ZS
                if cap:
                    ps[cap] = -1
                    cnt[1] -= 1
                    h2 ^= Z[cap][q]
                val = -search(1, depth - 1, -INF, -araw if araw > -INF else INF, 1, h2)
                if cap:
                    cnt[1] += 1
                    ps[cap] = q
                board[q] = cap
                board[p] = c
                ps[c] = p
                if abort[0]:
                    break
                done += 1
                adj = val - pen if val < WIN - 1000 else val
                vals.append((adj, m))
                if adj > best_adj:
                    best_adj = adj
                    cur_best = m

            if cur_best is not None and (not abort[0] or done >= 1):
                best_move = cur_best
                if not abort[0]:
                    vals.sort(key=lambda z: z[0], reverse=True)
                    order = [t_item[1] for t_item in vals]
                else:
                    order.remove(cur_best)
                    order.insert(0, cur_best)
            if abort[0]:
                break
            if best_adj > WIN - 1000 or best_adj < -WIN + 1000:
                break
            if time.process_time() > deadline or time.process_time() - start_time > 0.84:
                break

        c = best_move[1]
        p = best_move[2]
        q = best_move[3]
        cap = best_move[4]
        board[p] = 0
        board[q] = c
        ps[c] = q
        if cap:
            ps[cap] = -1
        key = tuple(ps)
        seen[key] = seen.get(key, 0) + 1
        ck = tuple(ps[ALLY_LO:ALLY_HI + 1])
        cfg[ck] = cfg.get(ck, 0) + 1
        return (pid_of[c], DIRN[q - p])

    try:
        res = decide()
    except Exception:
        # Nhánh dự phòng an toàn nếu gặp sự cố bất ngờ
        res = (0, 0)
        bestk = -INF
        dirs = ((-1, 1), (0, 1), (1, 1), (1, 0), (1, -1), (0, -1), (-1, -1), (-1, 0))
        for p in ally_list:
            for d in range(1, 9):
                dx, dy = dirs[d - 1]
                nx = p.x + dx
                ny = p.y + dy
                if nx < 1 or nx > W or ny < 1 or ny > H or (nx, ny) in walls_xy:
                    continue
                ok = True
                for a in ally_list:
                    if a.x == nx and a.y == ny:
                        ok = False
                for e in opp_list:
                    if e.x == nx and e.y == ny:
                        if not ((p.Type == 'R' and e.Type == 'S') or (p.Type == 'S' and e.Type == 'P') or (p.Type == 'P' and e.Type == 'R')):
                            ok = False
                if ok and ADV0[ny * STRIDE + nx] > bestk:
                    bestk = ADV0[ny * STRIDE + nx]
                    res = (p.pid, d)

    memo['used'] = used + (time.process_time() - start_time)
    return res

Success #stdin #stdout 0.07s 14392KB

stdin

copy

Standard input is empty

stdout

copy

Standard output is empty

https://ideone.com/sF489s

language:

Python 3 (python 3.12)

created:

visibility:

public

Share or Embed source code

Discover > Sphere Engine API

The brand new service which powers Ideone!

Discover > IDE Widget

Widget for compiling and running the source code in a web browser!

Discover > Sphere Engine API

Discover > IDE Widget

Choose your language