Problem 306: Paper-strip Game

(see projecteuler.net/problem=306)

The following game is a classic example of Combinatorial Game Theory:

Two players start with a strip of n white squares and they take alternate turns.
On each turn, a player picks two contiguous white squares and paints them black.
The first player who cannot make a move loses.

example

So, for 1 <= n <= 5, there are 3 values of n for which the first player can force a win.
Similarly, for 1 <= n <= 50, there are 40 values of n for which the first player can force a win.

For 1 <= n <= 1 000 000, how many values of n are there for which the first player can force a win?

My Algorithm

My first step was to write a simple brute-force routine (see bruteForce) which unfortuinately can't even solve the problem for n = 50 (but n < 28 is okay).
Due to the high number of won starting positions I only printed the lost positions: 1, 5, 9, 15, 21, 25
That sequence already exists in OEIS A215721 and has a promising name: "The values of N for which the 1xN domino-covering game is a second player win".

I found a short algorithm on that website:
nv(0) = 0
nv(1) = 0
nv(N+2) = least nonnegative integer not in {nimsum(nv(k),nv(N-k)) : k <= N}
(where nimsum(a,b) is the bitwise xor of a and b). The second player wins a game iff its nim-value is 0.
→ and pretty much the same is explained here: math.stackexchange.com/questions/615242/game-involving-tiling-a-1-by-n-board-with-1-x-2-tiles

The XOR operation already solved the Nim problem 301.
I store the XOR-ed value of a board size in my mex container (name stems from the StackExchange website, same as nv() in OEIS).

In order to analyze each board with 2,3,4,...,1000000 squares I perform these steps:

To improve performance I re-use found as much as possible and reset only those parts that were modified.
But no matter how hard I tried: the program was a bit slow and needed about 15 minutes to find the correct result.
I added a basic progress indicator and printed every 100th lost position: and all ended with a 7 !
Upon closer inspection I found that they were 680 apart each.
Then I printed every 10th lost solution and saw that they were 68 apart.
Then I printed every lost solution and needed a few second to see that a position n is lost if n - 34 was lost.
Unfortunately this pattern doesn't hold for the first 14 lost positions, therefore I hardcoded them.
See fast() for implementation which solves the problem so fast that I can't reliably measure execution time.
It keeps track of the most recent five values so that n % 5 refers to the index where the result of n - 5 was stored.

Alternative Approaches

fast() isn't the fastest algorithm: it still needs about 30000 loop iterations for n = 10^6.
I could exploit the very repetitive nature of the algorithm and avoid them - and maybe a super-smart compiler does exactly that.

The Sprague-Grundy algorithm can be used to solve a variety of similar problems (see blog.plover.com/math/sprague-grundy.html).

Note

If I had read OEIS A215721 more carefully I would have noticed that the "n - 34" pattern was described there ... oops.
Nevertheless, I felt quite proud of myself for a moment. Worth it.

Interactive test

You can submit your own input to my program and it will be instantly processed at my server:

Input data (separated by spaces or newlines):

This is equivalent to
echo 50 | ./306

Output:

(please click 'Go !')

Note: the original problem's input 1000000 cannot be entered
because just copying results is a soft skill reserved for idiots.

(this interactive test is still under development, computations will be aborted after one second)

My code

… was written in C++11 and can be compiled with G++, Clang++, Visual C++. You can download it, too.

#include <iostream>
#include <vector>
 
// uncomment to enable the slower algorithms
//#define BRUTEFORCE
//#define NIM_SUM
 
// return true if player to move wins on a board with numSquares (<= 64)
bool bruteForce(unsigned int numSquares, unsigned long long occupiedMask = 0)
{
// try each position and call yourself recursively
for (unsigned int pos = 0; pos + 1 < numSquares; pos++)
{
// two neighboring bits => 11 binary => 3
const unsigned long long twoSquares = 3ULL << pos;
// both bits empty ?
if ((occupiedMask & twoSquares) == 0)
// occupy both bits, if opponent loses then this is a winning move for the current play
if (!bruteForce(numSquares, occupiedMask | twoSquares))
return true;
}
 
// all moves are losing moves
return false;
}
 
// analyze all boards and count how often the first player wins
unsigned int nimSum(unsigned int maxSquares)
{
// https://oeis.org/A215721
// https://math.stackexchange.com/questions/615242/game-involving-tiling-a-1-by-n-board-with-1-x-2-tiles
// https://blog.plover.com/math/sprague-grundy.html
 
unsigned int numLost = 0;
 
std::vector<unsigned int> mex(maxSquares + 1, 0);
 
// find maximum value created by xoring states
unsigned int maxXor = 1;
while (maxXor < maxSquares)
maxXor <<= 1;
 
// track numbers of xored states
std::vector<bool> found(maxXor, false);
for (unsigned int i = 2; i <= maxSquares; i++)
{
// empty squares on the left and right side after the first move
int left = 0;
int right = i - 2;
 
// remember the highest XOR value
unsigned int lastXor = 0;
while (left <= right)
{
auto xored = mex[left] ^ mex[right];
found[xored] = true;
 
// remember the largest state (to speed up the clean-up/reset)
if (lastXor < xored)
lastXor = xored;
 
// slide one position to the right
left++;
right--;
}
 
// find smallest number which was NOT create during one of the XOR steps
unsigned int unused = 0;
while (found[unused])
unused++;
 
// finished computing current value
mex[i] = unused;
 
// first player only loses when mex[i] == 0
if (mex[i] == 0)
{
numLost++;
if (numLost % 100 == 0)
std::cout << i << "\t" << std::flush;
}
 
// erase all values (much faster than creating a new shiny std::vector)
for (unsigned int reset = 0; reset <= lastXor; reset++)
found[reset] = false;
}
 
return maxSquares - numLost;
}
 
// super-fast algorithm: implement the pattern observed in nimSum()
// nextLostPosition(n) = nextLostPosition(n - 5) + 34
// except for the first 14 values { 1, 5, 9, 15, 21, 25, 29, 35, 39, 43, 55, 59, 63 }
unsigned int fast(unsigned int maxSquares)
{
const auto Precomputed = 13;
const char Initial[Precomputed] = { 1, 5, 9, 15, 21, 25, 29, 35, 39, 43, 55, 59, 63 };
 
// store only the most recent 5 values
unsigned int last5[5];
 
unsigned int numLost = 0;
while (true)
{
// a(n) = a(n - 5) + 34
auto current = last5[numLost % 5] + 34;
// except for the first 13 values
if (numLost < Precomputed)
current = Initial[numLost];
 
// found enough
if (current > maxSquares)
return maxSquares - numLost;
 
// store the least recent 5 values
last5[numLost % 5] = current;
numLost++;
}
 
return 0; // never reached
}
 
int main()
{
unsigned int limit = 1000000;
std::cin >> limit;
 
// slowest algorithm
#ifdef BRUTEFORCE
unsigned int won = 0;
for (unsigned int n = 1; n <= limit; n++)
if (!bruteForce(n))
{
std::cout << n << " " << std::flush;
won++;
}
std::cout << won << std::endl;
#endif
 
// idea based on OEIS algorithm
#ifdef NIM_SUM
std::cout << nimSum(limit) << std::endl;
#endif
 
// my fastest algorithm
std::cout << fast (limit) << std::endl;
return 0;
}

This solution contains 26 empty lines, 34 comments and 6 preprocessor commands.

Benchmark

The correct solution to the original Project Euler problem was found in less than 0.01 seconds on an Intel® Core™ i7-2600K CPU @ 3.40GHz.
(compiled for x86_64 / Linux, GCC flags: -O3 -march=native -fno-exceptions -fno-rtti -std=gnu++11 -DORIGINAL)

See here for a comparison of all solutions.

Note: interactive tests run on a weaker (=slower) computer. Some interactive tests are compiled without -DORIGINAL.

Changelog

September 16, 2017 submitted solution
September 16, 2017 added comments

Difficulty

55% Project Euler ranks this problem at 55% (out of 100%).

Heatmap

Please click on a problem's number to open my solution to that problem:

green   solutions solve the original Project Euler problem and have a perfect score of 100% at Hackerrank, too
yellow solutions score less than 100% at Hackerrank (but still solve the original problem easily)
gray problems are already solved but I haven't published my solution yet
blue solutions are relevant for Project Euler only: there wasn't a Hackerrank version of it (at the time I solved it) or it differed too much
orange problems are solved but exceed the time limit of one minute or the memory limit of 256 MByte
red problems are not solved yet but I wrote a simulation to approximate the result or verified at least the given example - usually I sketched a few ideas, too
black problems are solved but access to the solution is blocked for a few days until the next problem is published
[new] the flashing problem is the one I solved most recently
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50
51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75
76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100
101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125
126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150
151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175
176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200
201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225
226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250
251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275
276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300
301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325
326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350
351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375
376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400
401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425
426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450
451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475
476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500
501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525
526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550
551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575
576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600
601 602 603 604 605 606 607 608 609 610 611 612 613 614 615
The 306 solved problems (that's level 12) had an average difficulty of 32.5% at Project Euler and
I scored 13526 points (out of 15700 possible points, top rank was 17 out of ≈60000 in August 2017) at Hackerrank's Project Euler+.

My username at Project Euler is stephanbrumme while it's stbrumme at Hackerrank.

Look at my progress and performance pages to get more details.

more about me can be found on my homepage, especially in my coding blog.
some names mentioned on this site may be trademarks of their respective owners.
thanks to the KaTeX team for their great typesetting library !