Cafe de La Regence (dedicated games at odds)

This forum is for general discussions and questions, including Collectors Corner and anything to do with Computer chess.

Moderators: Harvey Williamson, Steve B, Watchman

Forum rules
This textbox is used to restore diagrams posted with the fen tag before the upgrade.
Post Reply
Reinfeld
Member
Posts: 486
Joined: Thu Feb 17, 2011 3:54 am
Location: Tacoma, WA

Cafe de La Regence (dedicated games at odds)

Post by Reinfeld »

I'm starting this thread to explore a world I'm not sure dedicated matches have touched enough: games at odds, the old-school method of pitting players of unequal strength against each other.

This approach was the gambler's method in the old 19th-century chess clubs. While few books give it any sort of reference-level treatment, you can find a number of odds games in Staunton's "Chess Player's Companion" that amount to an opening survey. Other old masters such as Morphy played such games. The antiquarian openings manual, "Chess Openings Ancient and Modern," by Freeborough and Ranken (reprinted during the Fischer boom) includes MCO-style tables of ideal opening moves at odds.

More recently, in 2010, Steve sponsored a human v comp match where GM Larry Kaufman gave QR odds to Novag Constellation 3.6 and came away with a victory. There are other examples of modern GM-v-computer handicap matches.

This category offers wide latitude. While many options pitting machine-on-machine are available, it would be enormous fun to see forum members take a shot at human v machine matches. I promise to try this myself, and reveal the feeble depths of my chess ability.

The advent of clocks gradually superseded old-school odds games and replaced them with time odds (where the superior player gets less time.) This approach offers fairness in certain settings. I confess I'm not that interested in time odds here, though I won't condemn others who offer games on that basis.

I'm much more interested in other categories:

1. Traditional odds games
2. Feature handicaps

Traditional odds games are based on move and piece handicaps, ranging from pawn and move (up to pawn and three moves), then N, B, R and Q odds. For example, I would hope I could beat Super Connie 8 out of 10 times if it gave me odds of a knight, beer and moving whenever I feel like it (yes, I'll try it.)

Feature handicaps are more specific to dedicateds. They might include the following approaches - I'm sure other members can add to this list. Note that they can apply to human v machine as well as machine v machine:

1. Book on v Book off (Steve tried this in 2014 with a game between Rebell 5.0 book on v EAG V5 book off. The lower rated computer prevailed.) Link: http://www.hiarcs.net/forums/viewtopic.php?p=6767

2. Book off v Book off (obvious)

3. Book style v book style

4. Search type v search type (SEL v Brute Force, for example)

5. Search depth v search depth - some but not all dedicateds allow you to limit the horizon. A handicap of this nature might look like 2ply v 4ply.

6. Anything else you can think of.

Some, but not all of the questions this exercise might answer:

a. How important are opening books in terms of evaluating dedicated strength? (Steve noted that Rebell's win over EAG V5 represented an Elo difference of 225 points.)

b. What does brute force catch that selective search misses?

c. How badly do dedicateds screw up on a basic level when deprived of opening knowledge?

d. How much do you have to cripple a machine before you can beat it?

Have at it. I will try to post a game or two this weekend.

- R.
"You have, let us say, a promising politician, a rising artist that you wish to destroy. Dagger or bomb are archaic and unreliable - but teach him, inoculate him with chess."
– H.G. Wells
User avatar
Steve B
Site Admin
Posts: 10140
Joined: Sun Jul 29, 2007 10:02 am
Location: New York City USofA
Contact:

Post by Steve B »

Excellent new topic R
i shall follow the games with great interest

Thread Of The Year Regards
Steve
Reinfeld
Member
Posts: 486
Joined: Thu Feb 17, 2011 3:54 am
Location: Tacoma, WA

Post by Reinfeld »

As promised, a start to what I hope will be a fruitful thread.

I thought long and hard about the first odds game to get this going, and decided a comp v human contest would be fitting. But which comp?

I alluded to Super Connie when I started the thread, and I intend to go down that road eventually. That venerable, beloved machine is an ideal candidate, known to many of us - but I acquired it long after its heyday. Instead, I chose the first chess computer I ever owned, still functioning 32 years later: Fidelity Excellence. I believe I own the 3 mhz model, based on testing advice from Steve.

I don't play old Excellence as much as I used to, but I still drag it out now and then. Ugly, but so functional, so mean. For years, it beat me like a drum, sending me into frothing rage and flurries of vengeful takebacks (Oh, yeah? How do you like THIS, you sneaky bastard? Oh...oh...that move wasn't good either...well how about THIS...oh crap...god how I despise you.)

These days, I can hold my own for the most part, but as we all know, that Spracklen program is nasty. It punishes errors without mercy. My victories against it tend to be grinding, positional affairs. I almost never get away with tactical melees. So in the following game, you'll see me being conservative and careful.

Excellence was set to 30 seconds per move. I was set to as long as I wanted to to take. The game lasted about two hours. I left ponder ON, allowing Excellence to think on my time as compensation for the QN odds it gave me.

It should be obvious, but when you set the comp for odds play, you're also taking away its book by definition, throwing it on its own resources from the get-go. Anyway, here's the game, with light annotation:

[White "Reinfeld"]
[Black "Excellence"]
[Result "0-1"]

1.Nf3 d6

I finger-slipped, intending 1...d5, but figured it wouldn't matter. Decided to build a KID structure instead of slugging it out.

2.d4 Nf6 3.e3

[fen]rnbqkb1r/ppp1pppp/3p1n2/8/3P4/4PN2/PPP2PPP/R1BQKB1R w KQkq - 0 3[/fen]

Slightly surprising that Excellence entombed the dark-square bishop, but I'll take it.

g6 4.Bb5+ Nbd7 5.O-O Bg7 6.Bd2 O-O 7.Qe2 a6 8.Bd3 Re8 9.Ng5

[fen]r1bqr1k1/1ppnppbp/p2p1np1/6N1/3P4/3BP3/PPPBQPPP/R4RK1 w - - 3 9[/fen]

This was my longest think of the game, probably 20 minutes. The knight move from Excellence felt premature, and I couldn't figure out the intent. I debated whether to push the knight back with h6, but eventually decided that kicking out in the center with 9...e5 was faster. Obviously it's a threat to win a pawn, since White's queen is pinned, so Excellence protects the e-pawn and locks the d2 bishop into momentary passivity. Then push the knight back after that, force the trade, open up the board for my c8 bishop. So far, so good.

e5 10.c3 h6 11.Ne4 Nxe4 12.Bxe4 Nf6 13.Bc2 exd4 14.cxd4 d5

[fen]r1bqr1k1/1pp2pb1/p4npp/3p4/3P4/4P3/PPBBQPPP/R4RK1 w - - 0 15[/fen]

A bit slow on my part. Could have shoved the knight to e4 right away, but I was being cautious, giving Excellence too much respect.

15.f3 Bd7 16.Bd3 Rc8 17.e4

[fen]2rqr1k1/1ppb1pb1/p4npp/3p4/3PP3/3B1P2/PP1BQ1PP/R4RK1 w - - 0 17[/fen]

Not a good move from Excellence. Opening up the game here makes my job easier, due to the continuing pin on the e-file. The trades that follow just cement my advantage.

dxe4 18.fxe4 Nxe4 19.Bxe4 Bxd4+

[fen]2rqr1k1/1ppb1p2/p5pp/8/3bB3/8/PP1BQ1PP/R4RK1 w - - 0 20[/fen]

Apparently (per Stockfish) I should have played 19. ...Bf5, exploiting the pin, but I was obsessed with keeping the trades even, and I couldn't resist this check, which also attacks White's b2 pawn. Another series of trades follows - I'm not crushing it here, but I'm not losing my odds advantage, either.

20.Kh1 Bc6 21.Rf4 Bxe4 22.Rxe4 Rxe4 23.Qxe4 Bxb2 24.Rb1

[fen]2rq2k1/1pp2p2/p5pp/8/4Q3/8/Pb1B2PP/1R5K w - - 1 24[/fen]

Aaaahhhh. That mistake by Excellence loses a piece, though it took me a few minutes to see it. I was uneasy about 25. Qxb7, but the worry was needless. I failed to see the best followup move (25...Bc1, threatening mate), but that's OK.

Qxd2 25.Qxb7 Rd8 26.h3 Be5 27.Qxa6 Qf4 28.Kg1 Rd2

[fen]6k1/2p2p2/Q5pp/4b3/5q2/7P/P2r2P1/1R4K1 w - - 3 29[/fen]

It's all but over, assuming I don't screw it up. White has nothing left but spite checks. Always nice when Excellence starts responding instantly.

29.Qa8+ Kh7 30.Qh8+ Bxh8 31.Kh1 Qg3 32.Rg1 Be5 33.a4 Qh2# 0-1

Final position:

[fen]8/2p2p1k/6pp/4b3/P7/7P/3r2Pq/6RK w - - 1 34[/fen]

So there you go. Coffeehouse Excellence impudently tried to call me a knight-odds player. I collected my winnings, put on my gloves, and ordered a round for the onlookers.

I think my next effort will be comp on comp. Less stress regards,

- R.
"You have, let us say, a promising politician, a rising artist that you wish to destroy. Dagger or bomb are archaic and unreliable - but teach him, inoculate him with chess."
– H.G. Wells
User avatar
Steve B
Site Admin
Posts: 10140
Joined: Sun Jul 29, 2007 10:02 am
Location: New York City USofA
Contact:

Post by Steve B »

Reinfeld wrote: Excellence was set to 30 seconds per move. I was set to as long as I wanted to to take. I left ponder ON, allowing Excellence to think on my time as compensation for the QN odds it gave me.
An Unusal Form of Odds Game
Excellence got QN odds and you got time odds to compensate
dont recall seeing an odds game where an attempt is made to offset the odds given

Still.. its an interesting idea

Odd Odds Game Regard
Steve
Reinfeld
Member
Posts: 486
Joined: Thu Feb 17, 2011 3:54 am
Location: Tacoma, WA

Post by Reinfeld »

Here's my first attempt at a comp-v-comp odds game. The combatants:

1. Millennium Chess Genius Pro
2. Novag Super Constellation

Time control: 30s/move avg
Odds: Remove White's QN

Viewable here: https://lichess.org/CwSIB0HF#0

By wiki-elo measures, these machines are about 550 points apart. So the question becomes, how much is a knight worth under the circumstances?

Super Connie entered the tournament hall backed by an aging entourage of '80s hair band members and cheerleaders clad in flimsy pastels. MCGP sat impassively, waiting.

It's possible, as the experts know, to win an odds game by the simple strategy of trading down whenever possible, and turning the knight advantage into a durable endgame grind. However, endgames are a known weakness of old Super Connie, and the years have not been kind.

The game plays out in predictable fashion, but it turns out that MCGP is more eager to trade down than Super Connie. This results in two sets of doubled pawns for black, allowing MCGP to pick them off at leisure. At move 27, Super Connie is backed into a desperate situation, and squanders much of its advantage. Even then, the game might have been saved, but a howler at move 46 spells doom, and MCGP cruises home with a simple win.

[Date "2018.02.03"]
[Round "?"]
[White "Millennium CGP"]
[Black "Super Constellation"]
[Result "1-0"]
[Variant "From Position"]

1. e4 Nc6 2. Nf3 e5 3. Bb5 Nf6 4. d3 d6 5. O-O Be7 6. d4 O-O 7. Re1 Bg4 8. Bxc6 bxc6 9. dxe5 dxe5 10. Qxd8 Rfxd8 11. Nxe5 Bd7 12. Bg5 h6 13. Bf4 Bd6 14. Nxd7 Bxf4 15. Nxf6+ gxf6 16. Re2 Rab8 17. b3 Kg7 18. g3 Be5 19. Rf1 Bc3 20. f4 Rd6 21. Rf3 Rd2 22. Rxd2 Bxd2 23. Rd3 Bb4 24. Kf2 Re8 25. Kf3 Kg6 26. Rd4 a5 27. Rc4 c5 28. c3 Ba3 29. Ra4 Bb2 30. Rxa5 c4 31. bxc4 Bxc3 32. Ra7 Rd8 33. Rxc7 Ra8 34. a4 Rxa4 35. Ke3 Ra2 36. h3 Be1 37. f5+ Kg7 38. g4 Rh2 39. Kd4 Rxh3 40. c5 Bf2+ 41. Kd5 Rd3+ 42. Kc4 Rd4+ 43. Kc3 Rxe4 44. c6 Rxg4 45. Re7 Bd4+ 46. Kd3

[fen]8/4Rpk1/2P2p1p/5P2/3b2r1/3K4/8/8 w - - 3 46[/fen]

This is the losing moment. Black should have played 46...Ba1. The actual move allows a straightforward rook sac that decides the game.

Be5?? 47. Rxe5 Rg3+ 48. Re3 Rg1 49. Kc4 Kh7 50. c7 Rg8 51. Kb5 Rc8 52. Kc6 Kg8 53. Kd7 Rxc7+ 54. Kxc7 Kh7 55. Rh3 Kg8 56. Kd6 Kh7 57. Ke7 Kg7 58. Rg3+ Kh7 59. Kxf7 h5 60. Kxf6 Kh6 61. Rh3 Kh7 62. Rxh5+ Kg8 63. Rh6 Kf8 64. Rh8# 1-0

Super Connie left the arena, certain it could prevail at rook odds, if nothing else.

- R.
"You have, let us say, a promising politician, a rising artist that you wish to destroy. Dagger or bomb are archaic and unreliable - but teach him, inoculate him with chess."
– H.G. Wells
User avatar
Steve B
Site Admin
Posts: 10140
Joined: Sun Jul 29, 2007 10:02 am
Location: New York City USofA
Contact:

Post by Steve B »

Reinfeld wrote:Here's my first attempt at a comp-v-comp odds game. The combatants:

1. Millennium Chess Genius Pro
2. Novag Super Constellation

Time control: 30s/move avg
Odds: Remove White's QN


it turns out that MCGP is more eager to trade down than Super Connie. .
your comment about the MCGP trading down while giving piece odds is spot on..exactly the worst thing to do
I found this to be the case in the odds games I played (posted in the "Return of The Savant" Thread)
I used a 1 Min/Avg. TC
I had the MCGE give Knight Odds to the Savant(rated about 1500)
it scored 1.5-.5
I also tried R odds and the Savant won both games

The Return of the Golden Age Of Chess Regards
Steve
Reinfeld
Member
Posts: 486
Joined: Thu Feb 17, 2011 3:54 am
Location: Tacoma, WA

Post by Reinfeld »

And here's the next odds game. Since Super Connie failed so miserably with Knight odds against Millennium CGP, I figured I'd move up the rankings a bit, and see if Fidelity Chesster Challenger could come away with a win against MCGP at the same odds.

Chesster Challenger is a Par Excellence with a voice box. By rating at least, it's supposed to be 70 to 100 points stronger than Super Connie, and thus about 450 points weaker than MCGP. I was curious about how the Spracklen-based program would perform compared to Kittinger.

As before, the time controls were 30s/avg.

Fast run through here:

https://lichess.org/EDTXaas4

Much to my surprise, Chesster gave a poor account of himself, and played a poorer game than Super Connie by my reckoning. The knights and queens are off the board by move 10. Chesster's first error comes at move 12, when he falls into what I'll call the Fischer Trap, referring to Fischer's inexplicable blunder at move 29 in the first game of the famous 1972 match, when he snapped off a rook pawn with his bishop, and was forced to give it away in a head-slapping manner familiar to anyone who's played a club game.

The loss of the bishop reduces Chesster's advantage. MCGP pulls even by move 24. Soon, Chesster loses one rook, then pitches the game away at move 34, eventually resigning with a cheery sigh: "OK, you won this one -- I resign!"

[Date "2018.02.04"]
[White "Millennium CGP"]
[Black "Chesster Challenger"]

1. e4 d6 2. Be2 Nf6 3. d3 Nc6 4. Nf3 e5 5. Bg5 Be6 6. d4 exd4 7. Nxd4 Nxd4 8. Qxd4 h6 9. Bxf6 Qxf6 10. Qxf6 gxf6 11. O-O-O Rg8 12. g3 Bxa2

[fen]r3kbr1/ppp2p2/3p1p1p/8/4P3/6P1/bPP1BP1P/2KR3R w q - 0 13[/fen]

There's the Fischer Trap. Just pawn-grabbing that doesn't help.

13. b3 f5 14. Kb2 Bxb3 15. Kxb3 fxe4 16. f3 f5 17. Rhe1 Rg5 18. Bc4 a5 19. fxe4 a4+ 20. Ka2 Be7 21. Rb1 Ra7 22. e5 dxe5 23. Rxe5

[fen]4k3/rpp1b3/7p/4Rpr1/p1B5/6P1/K1P4P/1R6 w - - 0 23[/fen]

"Don't you think I'm great?" Chesster asks. His next move is not so great.

c6 24. Bd3 b5 25. Re6 Kd7 26. Rxh6 Ra8 27. Rd1 Kc7 28. Rh7 Re8 29. Rf1 Kd6 30. Bxf5 Rgg8 31. Rd1+ Kc5 32. Bd7 Bf6 33. Bxe8 Rxe8 34. Rh5+

[fen]4r3/8/2p2b2/1pk4R/p7/6P1/K1P4P/3R4 w - - 1 34[/fen]

"Come on, you love me, don't you?" Chesster says, before making the move that throws the game away for good.

Re5 35. Rxe5+ Bxe5 36. Rd3 Kb4 37. h4 Bf6 38. h5 Bg5 39. Rd6 Kc5 40. Rg6 Be3 41. h6 Bd4 42. h7 b4 43. Rg5+ Kd6 44. Rg4 b3+ 45. cxb3 axb3+ 46. Kxb3 Be5 47. Rg8 c5 48. g4 Kc6 49. g5

[fen]6R1/7P/2k5/2p1b1P1/8/1K6/8/8 w - - 0 49[/fen]

"Hey, you're getting good!" Well, yes. He resigned on the next move.

Kd5 50. g6 c4+ *

Hmm. Well, I knew MCGP was a strong machine. It won't be the only machine I use for these odds matchups, but it's handy, portable and new, so I'm interesting in running it through its paces, at least until I find a lower-level comp that can beat it with an extra knight.

Golden age patzers regards,

- R.
"You have, let us say, a promising politician, a rising artist that you wish to destroy. Dagger or bomb are archaic and unreliable - but teach him, inoculate him with chess."
– H.G. Wells
Reinfeld
Member
Posts: 486
Joined: Thu Feb 17, 2011 3:54 am
Location: Tacoma, WA

Post by Reinfeld »

After watching Super Connie and Chesster Challenger fail to bring home the point in knight-odds games against Millennium Chess Genius Pro, I decided to kick up the strength a bit and find a machine that would do the job properly.

My candidate: Saitek GK 2100.

This machine holds a curious place in our hobby, perhaps because it's so common and so copied, like a hit song that becomes tiresome. Yet I suspect we all view this aging comp with grudging respect. Like it or not, it set a standard in its time, combining a somewhat reasonable price point with an agreeable feature set and undeniable strength. The GK 2100 is a serious machine. It still plays a solid game. I've prevailed against lesser comps on occasion, but I don't think I've ever won a game against this thing without a takeback or two.

For my money, GK 2100 still marks what we might call the average line - the point when good comps become great. It hovers at the magical 2000 elo barrier, joined by Novag Obsidian (weaker) and Mephisto Milano, a more well-rounded romantic device, which is stronger than GK 2100 in my opinion (despite a slightly lower rating).

But enough of that. For this game against MCGP, I set the parameters at the same levels: 30s/move avg, selective search, random off.

By wiki-elo measures, GK 2100 stands at 2021 - about 280 points weaker than MCGP. The game result makes me wonder if elo points and centipawns have a certain equivalence - GK 2100 won, and never relinquished its advantage.

The game is rather long. Not so important. The opening, however, is very important. MCGP opened with 1. e4, and GK 2100, driven by its Morsch program, absent an opening book, seems to sense that the old Center Counter Defense, normally iffy due to White's tempo-gaining 3. Nc3, looks a whole lot better when the annoying White knight isn't there.

With that in mind, note the clever maneuvers by GK 2100 to preserve the black pawns as MCGP fishes for a blunder. This game is a triumph for Morsch's witty programming. See below.

Fast runthrough here: https://lichess.org/ELJiXcyA

Game score:

[White "Millennium CGP"]
[Black "GK 2100"]
[Result "0-1"]

1. e4 d5 2. exd5 Qxd5

[fen]rnb1kbnr/ppp1pppp/8/3q4/8/8/PPPP1PPP/R1BQKBNR w KQkq - 0 3[/fen]

Think of how many games you've played or viewed with this defense, how 3. Nc3 is automatic. Then think about the position when the knight is missing. So different.

GK 2100 plays nicely throughout from here, just prosecuting, forcing MCGP to waste time.


3. Ne2 Bf5 4. Ng3 Nc6 5. Nxf5 Qxf5 6. Bd3 Qe5+ 7. Qe2 O-O-O 8. Be4 Nf6 9. Bxc6 Qxe2+ 10. Kxe2 bxc6

[fen]2kr1b1r/p1p1pppp/2p2n2/8/8/8/PPPPKPPP/R1B4R w - - 0 11[/fen]

Fascinating. The queens are off the board. White has lost castling privileges, his king is exposed, and none of his pieces are developed.

11. d3 g6 12. Be3 a6 13. Rhe1 Bg7 14. Bd2 h5 15. Bc3 Rd5 16. Kf1 Kd7 17. Re2 Rd8 18. Rae1 e6 19. Re3 Ne8 20. Rf3 Ke7 21. Bxg7 Nxg7

[fen]3r4/2p1kpn1/p1p1p1p1/3r3p/8/3P1R2/PPP2PPP/4RK2 w - - 0 22[/fen]

Now White's bishop is gone. Black's rooks are doubled on a central file, and White's position is awkward.


22. Rf4 Rc5 23. Re2 Nf5 24. Ra4 Nd4 25. Re4 Nxc2 26. Rec4 Rxc4 27. Rxc4 Nd4

[fen]3r4/2p1kp2/p1p1p1p1/7p/2Rn4/3P4/PP3PPP/5K2 w - - 1 28[/fen]

White is down to one rook. All he can do is tack back and forth and try to pick off pawns. Black's centralized King is more than capable of protecting his soldiers.

28. Ra4 Nb5 29. Rxa6 Rxd3 30. Ke1 Nd4 31. Ra7 Kd6 32. Ra8 Kd7 33. Rf8 Ke7 34. Rc8 Nc2+ 35. Ke2 Rd7 36. Ra8 c5 37. Ra5 Kd6 38. Ra6+ Ke5 39. Ra5 Kd5 40. Ra8 Ke4 41. f3+ Kd5 42. Ra7 f6 43. Ra8 c6

[fen]R7/3r4/2p1ppp1/2pk3p/8/5P2/PPn1K1PP/8 w - - 0 44[/fen]

Look at the position. Nothing White can do.

44. Kd2 Nd4 45. Rb8 Kc4 46. Ke3 Nc2+ 47. Ke2 Rd3 48. Ra8 Nd4+ 49. Ke1 Re3+ 50. Kd1 e5 51. h3 f5 52. Ra7 Rd3+ 53. Ke1 e4 54. fxe4 fxe4 55. Ra4+ Kb5 56. Ra8 e3 57. Re8 Kc4 58. Re4 Kd5 59. Re8 Rd2 60. Rxe3 Rxg2 61. Kd1 Rxb2 62. Rg3 Rxa2 63. Rxg6 h4 64. Rg5+ Kc4 65. Kc1 Ne2+ 66. Kd1 Nf4 67. Kc1 Kb3 68. Rg8 Nd3+ 69. Kd1 Kc3 70. Rg2 Ra1+ 71. Ke2

[fen]8/8/2p5/2p5/7p/2kn3P/4K1R1/r7 w - - 15 71[/fen]

Finally, the execution.

Nf4+ 72. Kf3 Nxg2 73. Kxg2 Ra3 74. Kf3 Kd4+ 75. Kg4 c4 76. Kf5 Rg3 77. Ke6 c3 78. Kf5 c2 79. Ke6 c1=Q 80. Kd7 Rg7+ 81. Kc8 Qb2 82. Kd8 Qb8# 0-1

You won't hear these words often on this site but,

all hail Morsch regards,

- R.

P.S.: No more MCGP games for the moment. The next series will focus on opening books. Stay tuned.
"You have, let us say, a promising politician, a rising artist that you wish to destroy. Dagger or bomb are archaic and unreliable - but teach him, inoculate him with chess."
– H.G. Wells
Post Reply