1. Joined
    24 May '08
    Moves
    717
    11 Sep '16 11:102 edits
    I won't bore anyone with the full analysis (unless they really want it then I can PM it) but I have analysed the following game with various different engines at various different settings. The only real anomaly is GPs top 1 match in the rather rapid analysis with Stockfish at max depth= 20 ply. All the rest are pretty consistent, given the odd swapping of move preference in a tiny sample size like this skews %s in a very big way.
    If anyone thinks Houdini 1.5a on a quad core at fixed depth of 20 ply or Deep Rybka 3 at 60 seconds per half-move is trivial analysis, then I think you should be trouncing Carlsen rather than wasting time posting on this forum!

    Game 11852811

    Houdini 1.5a x4 CPU 30s max depth 20 ply hash 512 Mb (analysis took 36 min to complete)
    { White: greenpawn34 }
    { Top 1 Match: 11/24 ( 45.8% )
    { Top 2 Match: 14/24 ( 58.3% )
    { Top 3 Match: 17/24 ( 70.8% )
    { Top 4 Match: 18/24 ( 75.0% )

    { Black: cenerentola }
    { Top 1 Match: 22/25 ( 88.0% )
    { Top 2 Match: 22/25 ( 88.0% )
    { Top 3 Match: 24/25 ( 96.0% )
    { Top 4 Match: 24/25 ( 96.0% )

    Houdini 1.5a x4 CPU fixed depth 20 ply hash 512 Mb (analysis took 1hr 14 min to complete)
    { greenpawn34 (Games: 1) }
    { Top 1 Match: 9/24 ( 37.5% )
    { Top 2 Match: 12/24 ( 50.0% )
    { Top 3 Match: 17/24 ( 70.8% )
    { Top 4 Match: 19/24 ( 79.2% )

    { cenerentola (Games: 1) }
    { Top 1 Match: 20/25 ( 80.0% )
    { Top 2 Match: 22/25 ( 88.0% )
    { Top 3 Match: 24/25 ( 96.0% )
    { Top 4 Match: 24/25 ( 96.0% )

    Stockfish 7 x64 popcnt 30s max depth 20 ply hash 512 Mb (analysis took 21 min to complete)
    { White: greenpawn34 }
    { Top 1 Match: 5/24 ( 20.8% )
    { Top 2 Match: 12/24 ( 50.0% )
    { Top 3 Match: 13/24 ( 54.2% )
    { Top 4 Match: 15/24 ( 62.5% )

    { Black: cenerentola }
    { Top 1 Match: 20/25 ( 80.0% )
    { Top 2 Match: 24/25 ( 96.0% )
    { Top 3 Match: 24/25 ( 96.0% )
    { Top 4 Match: 24/25 ( 96.0% )

    Stockfish 7 x64 popcnt 60s max depth 26 ply hash 512 Mb (analysis took 54 min to complete)
    { White: greenpawn34 }
    { Top 1 Match: 8/24 ( 33.3% )
    { Top 2 Match: 12/24 ( 50.0% )
    { Top 3 Match: 14/24 ( 58.3% )
    { Top 4 Match: 16/24 ( 66.7% )

    { Black: cenerentola }
    { Top 1 Match: 16/25 ( 64.0% )
    { Top 2 Match: 23/25 ( 92.0% )
    { Top 3 Match: 23/25 ( 92.0% )
    { Top 4 Match: 24/25 ( 96.0% )

    Deep Rybka 3_64 60s max depth 26 ply hash 512 Mb (analysis took 1hr 9 min to complete)
    { White: greenpawn34 }
    { Top 1 Match: 9/24 ( 37.5% )
    { Top 2 Match: 13/24 ( 54.2% )
    { Top 3 Match: 17/24 ( 70.8% )
    { Top 4 Match: 19/24 ( 79.2% )

    { Black: cenerentola }
    { Top 1 Match: 17/25 ( 68.0% )
    { Top 2 Match: 22/25 ( 88.0% )
    { Top 3 Match: 24/25 ( 96.0% )
    { Top 4 Match: 24/25 ( 96.0% )
  2. e4
    Joined
    06 May '08
    Moves
    42492
    11 Sep '16 13:34
    So now I'm down to 20% first move. My confidence is shattered...thank you all.

    I have an ongoing game v Scottish Geek and have no qualms about talking
    an ongoing game because I cannot see a way of bring up the reserves
    so am going to go Qh3+ Qg3+ Qh3+ etc and etc.. (I'm Black)





    I'm going to tell Scottish Geek to keep the going and not claim a draw.
    We will do this for 100 moves so I'll have 100 first choices and be good!

    Yes their are individuals who cheat but I can confirm the vast majority don't.
    Look at the blog, look at my games. Blunders galore.
  3. Joined
    30 Apr '16
    Moves
    1056
    12 Sep '16 02:011 edit
    Originally posted by Zygalski
    I won't bore anyone with the full analysis (unless they really want it then I can PM it) but I have analysed the following game with various different engines at various different settings. The only real anomaly is GPs top 1 match in the rather rapid analysis with Stockfish at max depth= 20 ply. All the rest are pretty consistent, given the odd swapping o ...[text shortened]...
    { Top 2 Match: 22/25 ( 88.0% )
    { Top 3 Match: 24/25 ( 96.0% )
    { Top 4 Match: 24/25 ( 96.0% )
    I was NOT suggesting that a 30s Houdini analysis is "trivial" or that a human player could ever beat it. That completely misses the point. A 30s Houdini analysis found a 5s Stockfish player to be very superior to a 3600s Stockfish player, which is obviously utter nonsense. Only 55% of the vastly superior player's moves were found to be "TOP 1" moves by 30s Houdini. A 30s analysis will miss many moves that a really good player (and a super human player such as my 3600s Stockfish) will find.

    I used a recent i7 processor configured to use 8 threads so in the 3600s analysis it typically checked 4-12 billion positions. I think this is a bit better than a 30s Houdini analysis that found 45% of the moves to not be best.
  4. Joined
    24 May '08
    Moves
    717
    12 Sep '16 18:124 edits
    Originally posted by SkippyJoe
    I was NOT suggesting that a 30s Houdini analysis is "trivial" or that a human player could ever beat it. That completely misses the point. A 30s Houdini analysis found a 5s Stockfish player to be very superior to a 3600s Stockfish player, which is obviously utter nonsense. Only 55% of the vastly superior player's moves were found to be "TOP 1" moves by 30s ...[text shortened]... ink this is a bit better than a 30s Houdini analysis that found 45% of the moves to not be best.
    But the method I used (and you criticised) is designed to catch idiotic engine users who simply play one of the top 4 best moves time after time in all their games. I never said anything about the moves being spat out of the engine after 30 seconds analysis as being the 'best' ones. You made that leap.

    These idiots play on here because they'd get slaughtered over on ICCF where the serious centaurs play. The idiot probably has 10+ games in progress so logs in, fires up the engine, picks top engine choice move after a few moments, plays move, next game - rinse & repeat. I very much doubt any RHP centaurs analyse individual moves overnight on massively powerful multi-core pcs. Somewhat a case of overkill if they did!

    What sound reason is there that an unknown online player should consistently match Houdini at 30 sec per half move at a rate massively higher than that of all Super GMs OTB & all pre 1990s World Correspondence Chess Championship finalists... one that doesn't involve using an engine to suggest moves?
  5. Joined
    30 Apr '16
    Moves
    1056
    13 Sep '16 17:02
    I am sorry, but although I admit that your conclusions may well be correct, it's clear to me that they are unjustified. You stated "What sound reason is there that an unknown online player should consistently match Houdini at 30 sec per half move at a rate massively higher than that of all Super GMs OTB & all pre 1990s World Correspondence Chess Championship finalists". The answer is clear from my analysis that showed that a far better player than 30s Houdini (being a 3600s Stockfish player) was considered by 30s Houdini to be a much weaker player by 30s Houdini than a 5s Stockfish player. Top players will find some of the better moves that 30s Houdini does not. This is clear given that 30s Houdini rated 45% of 3600s Stockfish's moves as not the best moves. I am NOT suggesting that any human would find all these 45% but some very good players will find some of these which you do not seem to accept.

    Honestly, I just think you are being unfair because people do not like the fact that there are others who are simply better than them. Nobody likes to lose, but Cenerentola is really good. I have played him multiple times (where I am using a different name) and I am pretty convinced, for what it is worth, that he is for real. He creates incredible complexity in games that I do not find from playing computers as I have done a lot. Some other 2400+ players that I have played on RHP do seem much more computer-like but I still do not buy your approach to assess this.

    My other comment is "Who cares?" We are all anonymous (well, especially me) and nothing is at stake other than some arbitrary rating on a website. I am only here to have fun. To be honest though, it's not fun for me to play cenerentola, but that's just because he is too good for me.
  6. Joined
    24 May '08
    Moves
    717
    13 Sep '16 17:184 edits
    Originally posted by SkippyJoe
    I am sorry, but although I admit that your conclusions may well be correct, it's clear to me that they are unjustified. You stated "What sound reason is there that an unknown online player should consistently match Houdini at 30 sec per half move at a rate massively higher than that of all Super GMs OTB & all pre 1990s World Correspondence Chess Championsh ...[text shortened]... these 45% but some very good players will find some of these which you do not seem to accept.
    Ok, but humans play human moves, make human positional choices whereas engine users play like engines.

    You can't have it both ways. If someone repeatedly plays like an engine & they're highly rated on this site, you seem to be suggesting that the moves aren't really engine like at all, since if the engine had longer to look at the position then it'd choose different moves, like a Super GM would.

    Now that poses a few problems. As I showed above, engines like certain moves in certain games regardless of how long you given them to look at positions.
    Second, where is your proof that GMs (and presumably super-humanly strong untitled & online anonymous unassisted players) would choose top engine choice moves more consistently if the engine is given far, far longer than 30 seconds to look at each position? Where is your evidence? It should be a fairly straightforward process for you to analyse all of say Fischer's games vs Spassky in 1972 at 30 seconds per half move & then re-analyse at 5 minutes per half move for top 4 choices. I await with baited breath....

    Also, how would you look for suspected cheats if you don't match the moves they make with those suggested by engines?

    Finally, why do Super Gms & OTB Correspondence Masters all match engines to
    around the same degree (within a few % points) when using this approach? You'd expect a few outliers, but if I analyse 20 games from Capablanca or Carlsen, I'd expect roughly the same engine match rates.

    A couple of years ago I analysed the top 10 Super GM games; 20 most recently completed for each player (2014/2013) from www.chessgames.com which all have a minimum of 35 total moves vs 2500+ rated opponents. I ignored any games which had rapidplay/blitz in the pgn title. Moves that appeared in the 4.3m game database were excluded from analysis.

    { Magnus Carlsen (Games: 20) }
    { Top 1 Match: 477/828 ( 57.6% ) Opponents: 452/821 ( 55.1% )
    { Top 2 Match: 617/828 ( 74.5% ) Opponents: 594/821 ( 72.4% )
    { Top 3 Match: 690/828 ( 83.3% ) Opponents: 667/821 ( 81.2% )
    { Top 4 Match: 732/828 ( 88.4% ) Opponents: 709/821 ( 86.4% )

    { All Players }
    { Top 1 Match: 929/1649 ( 56.3% )
    { Top 2 Match: 1211/1649 ( 73.4% )
    { Top 3 Match: 1357/1649 ( 82.3% )
    { Top 4 Match: 1441/1649 ( 87.4% )


    { Levon Aronian (Games: 20) }
    { Top 1 Match: 483/814 ( 59.3% ) Opponents: 448/808 ( 55.4% )
    { Top 2 Match: 634/814 ( 77.9% ) Opponents: 601/808 ( 74.4% )
    { Top 3 Match: 699/814 ( 85.9% ) Opponents: 671/808 ( 83.0% )
    { Top 4 Match: 730/814 ( 89.7% ) Opponents: 723/808 ( 89.5% )

    { All Players }
    { Top 1 Match: 931/1622 ( 57.4% )
    { Top 2 Match: 1235/1622 ( 76.1% )
    { Top 3 Match: 1370/1622 ( 84.5% )
    { Top 4 Match: 1453/1622 ( 89.6% )


    { Vladimir Kramnik (Games: 20) }
    { Top 1 Match: 535/938 ( 57.0% ) Opponents: 523/936 ( 55.9% )
    { Top 2 Match: 701/938 ( 74.7% ) Opponents: 720/936 ( 76.9% )
    { Top 3 Match: 776/938 ( 82.7% ) Opponents: 785/936 ( 83.9% )
    { Top 4 Match: 819/938 ( 87.3% ) Opponents: 826/936 ( 88.2% )

    { All Players }
    { Top 1 Match: 1058/1874 ( 56.5% )
    { Top 2 Match: 1421/1874 ( 75.8% )
    { Top 3 Match: 1561/1874 ( 83.3% )
    { Top 4 Match: 1645/1874 ( 87.8% )


    { Veselin Topalov (Games: 20) }
    { Top 1 Match: 466/853 ( 54.6% ) Opponents: 465/848 ( 54.8% )
    { Top 2 Match: 642/853 ( 75.3% ) Opponents: 626/848 ( 73.8% )
    { Top 3 Match: 728/853 ( 85.3% ) Opponents: 694/848 ( 81.8% )
    { Top 4 Match: 772/853 ( 90.5% ) Opponents: 742/848 ( 87.5% )

    { All Players }
    { Top 1 Match: 931/1701 ( 54.7% )
    { Top 2 Match: 1268/1701 ( 74.5% )
    { Top 3 Match: 1422/1701 ( 83.6% )
    { Top 4 Match: 1514/1701 ( 89.0% )


    { Fabiano Caruana (Games: 20) }
    { Top 1 Match: 560/1018 ( 55.0% ) Opponents: 508/1019 ( 49.9% )
    { Top 2 Match: 749/1018 ( 73.6% ) Opponents: 694/1019 ( 68.1% )
    { Top 3 Match: 834/1018 ( 81.9% ) Opponents: 802/1019 ( 78.7% )
    { Top 4 Match: 887/1018 ( 87.1% ) Opponents: 854/1019 ( 83.8% )

    { All Players }
    { Top 1 Match: 1068/2037 ( 52.4% )
    { Top 2 Match: 1443/2037 ( 70.8% )
    { Top 3 Match: 1636/2037 ( 80.3% )
    { Top 4 Match: 1741/2037 ( 85.5% )


    { Alexander Grischuk (Games: 20) }
    { Top 1 Match: 479/889 ( 53.9% ) Opponents: 471/891 ( 52.9% )
    { Top 2 Match: 636/889 ( 71.5% ) Opponents: 635/891 ( 71.3% )
    { Top 3 Match: 707/889 ( 79.5% ) Opponents: 726/891 ( 81.5% )
    { Top 4 Match: 765/889 ( 86.1% ) Opponents: 768/891 ( 86.2% )

    { All Players }
    { Top 1 Match: 950/1780 ( 53.4% )
    { Top 2 Match: 1271/1780 ( 71.4% )
    { Top 3 Match: 1433/1780 ( 80.5% )
    { Top 4 Match: 1533/1780 ( 86.1% )


    { Hikaru Nakamura (Games: 20) }
    { Top 1 Match: 504/871 ( 57.9% ) Opponents: 441/869 ( 50.7% )
    { Top 2 Match: 641/871 ( 73.6% ) Opponents: 624/869 ( 71.8% )
    { Top 3 Match: 708/871 ( 81.3% ) Opponents: 700/869 ( 80.6% )
    { Top 4 Match: 764/871 ( 87.7% ) Opponents: 752/869 ( 86.5% )

    { All Players }
    { Top 1 Match: 945/1740 ( 54.3% )
    { Top 2 Match: 1265/1740 ( 72.7% )
    { Top 3 Match: 1408/1740 ( 80.9% )
    { Top 4 Match: 1516/1740 ( 87.1% )


    { Viswanathan Anand (Games: 20) }
    { Top 1 Match: 439/732 ( 60.0% ) Opponents: 401/732 ( 54.8% )
    { Top 2 Match: 563/732 ( 76.9% ) Opponents: 539/732 ( 73.6% )
    { Top 3 Match: 627/732 ( 85.7% ) Opponents: 607/732 ( 82.9% )
    { Top 4 Match: 652/732 ( 89.1% ) Opponents: 641/732 ( 87.6% )

    { All Players }
    { Top 1 Match: 840/1464 ( 57.4% )
    { Top 2 Match: 1102/1464 ( 75.3% )
    { Top 3 Match: 1234/1464 ( 84.3% )
    { Top 4 Match: 1293/1464 ( 88.3% )


    { Sergey Karjakin (Games: 20) }
    { Top 1 Match: 460/876 ( 52.5% ) Opponents: 471/876 ( 53.8% )
    { Top 2 Match: 621/876 ( 70.9% ) Opponents: 630/876 ( 71.9% )
    { Top 3 Match: 715/876 ( 81.6% ) Opponents: 708/876 ( 80.8% )
    { Top 4 Match: 763/876 ( 87.1% ) Opponents: 754/876 ( 86.1% )

    { All Players }
    { Top 1 Match: 931/1752 ( 53.1% )
    { Top 2 Match: 1251/1752 ( 71.4% )
    { Top 3 Match: 1423/1752 ( 81.2% )
    { Top 4 Match: 1517/1752 ( 86.6% )


    { Maxime Vachier-Lagrave (Games: 20) }
    { Top 1 Match: 475/801 ( 59.3% ) Opponents: 457/797 ( 57.3% )
    { Top 2 Match: 613/801 ( 76.5% ) Opponents: 583/797 ( 73.1% )
    { Top 3 Match: 685/801 ( 85.5% ) Opponents: 656/797 ( 82.3% )
    { Top 4 Match: 720/801 ( 89.9% ) Opponents: 701/797 ( 88.0% )

    { All Players }
    { Top 1 Match: 932/1598 ( 58.3% )
    { Top 2 Match: 1196/1598 ( 74.8% )
    { Top 3 Match: 1341/1598 ( 83.9% )
    { Top 4 Match: 1421/1598 ( 88.9% )


    { Top 10 FIDE Super GMs (Games: 200) }
    { Top 1 Match: 4878/8620 ( 56.6% )
    { Top 2 Match: 6417/8620 ( 74.4% )
    { Top 3 Match: 7169/8620 ( 83.1% )
    { Top 4 Match: 7604/8620 ( 88.2% )
  7. Joined
    30 Apr '16
    Moves
    1056
    13 Sep '16 17:291 edit
    Interesting and potentially good arguments. I will need to consider them and get back to you. BTW I am not saying you are wrong - I just do not buy how you justify your conclusions especially given my several encounters with cenerentola, who has impressed me (but I am not a master level player).
  8. Joined
    05 Nov '08
    Moves
    13417
    15 Sep '16 10:59
    Originally posted by SkippyJoe
    I am sorry, but although I admit that your conclusions may well be correct, it's clear to me that they are unjustified. You stated "What sound reason is there that an unknown online player should consistently match Houdini at 30 sec per half move at a rate massively higher than that of all Super GMs OTB & all pre 1990s World Correspondence Chess Championsh ...[text shortened]... though, it's not fun for me to play cenerentola, but that's just because he is too good for me.
    if you are using a different name, isn't that multiple accounts and a bannable offence ?
  9. e4
    Joined
    06 May '08
    Moves
    42492
    15 Sep '16 23:50
    I think it's OK to have joint account. Not too sure why but as long as
    you only make moves with one account then it's OK. (I think).
  10. Joined
    05 Nov '08
    Moves
    13417
    16 Sep '16 06:05
    ToS 3a basically says you will not create more than one account.
  11. Subscribersonhouse
    Fast and Curious
    slatington, pa, usa
    Joined
    28 Dec '04
    Moves
    53223
    16 Sep '16 17:48
    Originally posted by st40
    ToS 3a basically says you will not create more than one account.
    A lot of things formerly bannable are let go with a wink these days.
  12. Joined
    05 Nov '08
    Moves
    13417
    19 Sep '16 12:08
    btw
    I think SkippyJoe and Centerola might just be the same person....
  13. Joined
    30 Apr '16
    Moves
    1056
    19 Sep '16 17:26
    Originally posted by st40
    btw
    I think SkippyJoe and Centerola might just be the same person....
    OMG, you caught me! The shame! the shame!

    You people are sad. There are many very strong players. You need to accept that some of them will play on this site and beat you.
  14. Standard memberBigDogg
    Secret RHP coder
    on the payroll
    Joined
    26 Nov '04
    Moves
    155080
    24 Sep '16 12:46
    Originally posted by SkippyJoe
    OMG, you caught me! The shame! the shame!

    You people are sad. There are many very strong players. You need to accept that some of them will play on this site and beat you.
    Which would you say there are more of - strong human players, or engines stronger than human players?

    Do you acknowledge that some humans are tempted to cheat by using chess engines in their games?

    OK, so st40 overstepped in claiming you are someone you're not. But the frustration of playing engine users makes some people [understandably] paranoid.
  15. Standard memberDeepThought
    Losing the Thread
    Quarantined World
    Joined
    27 Oct '04
    Moves
    87415
    24 Sep '16 13:10
    Originally posted by sonhouse
    A lot of things formerly bannable are let go with a wink these days.
    This was discussed about a decade ago when games moderation was first introduced. According to the discussion then the site administrators will not act unless the accounts are being misused in some way. The purpose being to prevent this kind of thing: @Golden-King
Back to Top

Cookies help us deliver our Services. By using our Services or clicking I agree, you agree to our use of cookies. Learn More.I Agree