Russell Sherwood

Russell Sherwood  Wednesday, December 13, 2017

I get asked with tiresome regularity “Which engine is best for CC?” My general answer to this is “none of them” as all the main engines are written and tested with fairly short time controls in mind. As I have written in the past some are better than others and most can be bent into playing better for CC.

 

These are a number of engines with modifications to aid CC players – CorChess and ThinksFish come to mind but these only tend to look at modifying the search rather than the many other areas which could be improved for CC players.

 

Recently I was working on my publication on Engine Analysis techniques (over 30 now!) and what struck me is that most of these methods have to be used to overcome the “Blitz” nature of Chess Engine Settings. So the next thought that popped into my head was “Why not write one for CC”? I mulled this one over for a while and considered the pro and cons.

 

Most things fell into place but one issue remains – which is testing. If I want an engine for CC, then it needs to be tested in CC Conditions as the engine will almost certainly be weaker in Blitz Conditions (as want a CC engine to find a better move in a slower way, rather than a good move quickly). This makes testing rather difficult unless a few people are involved.

 

My first objective is to create an ASMFish/ CorChess Hybrid prior to the development of a CC specialist engine.

 

Any player interested in getting involved (not just Welsh players for once!) please get in touch. You don’t need to be a programmer, Ideas and testing are just as important!

 

Russell Sherwood  Wednesday, December 13, 2017

World Championship Preliminaries

 

The ICCF World Championship Cycle starts again in 2018 with the Preliminaries. It costs £25 to enter (if paid via the WCCF), if qualified and there are two routes to qualification.

 

1. Winning a Master Class Promotion Tournament. Not is first place is shared this leads to a fractional qualification.
2. Federation Nomination. We have a very limited number of nominations and so invite applications from interested Welsh Players with a rating above 2250 on 2018/1 rating list. If interested please get in touch indicating why you think you should obtain one of the nominations.

 

 

To explain why the rating limit of 2250 is used. These events a highly competitive and it is unlikely that any player with a rating below this level will have a reasonable chance of Advancement (which is the primary objective of the entry, not Norms or Rating points)

 

Russell Sherwood  Wednesday, December 13, 2017

Most CC players have heard of Dead Man’s Defence, the unsavoury technique used to put off the result of the game (generally a loss for the person using DMD). Here I want to discuss it’s cousin, Dead Man’s Attack!

 

Picture the situation. You are playing a game, which is into the early endgame. Engine Analysis shows a nice +1 core in your favour but neither you nor the engine knows how to make progress.  What then tends to happen is that the moves played shuffle the pieces around without making any visible progress and after a little while longer a draw is agreed as its one of the positions the engine does not understand. It tends to be in practice that a Fortress was in place which secured the position.

 

That is what is supposed to happen……in a minority of games we see the “winning” player refusing the give up his “win” and continue shuffling. Now I am on the wrong side of one of these games at the moment and am at move 110! An interesting fact for me is that my opponent will not lose any rating points for a draw which is often the motivation to keep shuffling, hoping your opponent makes a mistake.

 

This piqued my interest, how often does this take place? Using my trusty Chessbase I took a look at the longest ICCF games in the database and the results are interesting……..

 

137 games more than 120 moves long

56 of them since 2012

33 of these turned into draws, leaving 23 as decisive

 

So the message to those playing DMA is quite clear – don’t its chance of success is less than 50%

Russell Sherwood  Monday, December 11, 2017

Most CC players tend to be aware that engines don’t play certain types of positions very well and traditionally the King’s Indian Defence has been seen as an area where Engines are close to clueless!

 

Fewer players tend to also know which Openings/Positions engines tend to do well or not so well in but very few have a comprehensive view of this.

 

This does not have to be the case. Some very interesting research has been taking place at  http://www.amateurschach.de/ One that is of interest to ambitious CC players is his project to create a Openings rating list of sorts.

 

There is a great description of what has been done but in simple terms, various engines have been tested and the results then tabulated by ECO Code.  At the Summary level Stockfish/Houdini and Komodo are at the top but if you into individual ECO codes or groups of codes, it shows that all of these engines are perform well/badly in certain positions!  Now this information can be used in a number of ways…..

• You can make sure your engine/opening combination matches up well (which one you change is your choice!)
• If you know which engine your opponent uses as their primary analysis tool, then you could steer the game into an area that engine does badly.
• If you use an analysis technique involving multiple engines, you will know which ones tend to be more accurate in which type of position.

Is this a game changer? No, but when you consider that the difference in performance between the “Big 3” Engines is 15% in certain positions it does appear to be a potential way to gain a nice “edge”

Russell Sherwood  Sunday, December 10, 2017

Alpha Go a GoGo

 

https://arxiv.org/pdf/1712.01815.pdf

 

Earlier this week an interesting paper was published regarding Google Deepmind’s Alpha Go Zero teams research into Chess. Since published this report and various sensationalist articles have sent seismic waves through the Chess community.

The basic claim is that this program learned to play chess in 3 hours to the point that it beat Stockfish 28 Wins  - 72 Draws – 0 Defeats.

Whilst, without a doubt, this approach will eventually become the prevalent approach to computer chess, the reports circulating are a little over the top.

Looking more closely at the details. The 3 hours taken to reach these levels was on hardware so specialised and new that it is only available to Google’s Deepmind team. For us mere mortals this can be thought of as a Supercomputer. So whilst 3 hours is very impressive, if it had been on typical hardware this would have taken much longer.

The conditions of the match were somewhat in Alpha Go Zero’s favour. Firstly whilst Stockfish ran on 64 Cores, it only had 1GB of HashTable available, probably around 1% of what is required. This is the equivalent of putting a Ferrari engine into a Mini!

The second issue is the time control used – 1 minute per move. All of the top Chess engines have very efficient algorithms to manage time and identify when to spend more time on a move or when to spend less – very similar to Human behaviour.

The third difference and probably the most telling, is that StockFish played without the benefit of an opening book. A fair event would have been either an opening book or a Brainfish/Cerebellum combination.

The belief amongst the Computer Chess enthusiasts is that if the conditions above had been “fair” the result would have been much closer.

In addition, for a wider test, the same match should take place with Houdini, Komodo.

So to summarise, this approach is the path of the future but the current “wow factor” is probably premature. For those interested, the Giraffe engine is worth looking at and some projects which can be seen on Github putting AlphaGo Zero’s ideas into practice!

Russell Sherwood  Sunday, December 10, 2017

Where there’s a way there’s a will!

 

At first glance, you may think I have got this saying the wrong way round! In typical use, you would be correct but in the world of improvement, this is a more common saying. In Chess (as in other things) we all want to improve and there is a belief among many that simply wanting to improve (so-called positive thinking) is enough to get better on its own. However, the evidence tends to suggest otherwise and that having a clear path to improvement is more important.

 

So how do we get a clear path to improvement In CC?  Much of the advice that does float around is from high rated players who started out highly rated (i.e. their initial rating was very high and stayed there) and is so not necessarily relevant to a typical players journey.

 

Where do we want our advice to come from? Preferably from people who have undergone the journey we wish to undertake, i.e. who have seen their ratings improve over an extended period.  Downloading a few ICCF Rating lists and some manipulation in Excel leads to the list below.  

RUS		Kornev, Aleksey Nikolaevich	12	M	2468
USA		Risquet, Carmelo	12	M	2386
ENG		Stebbings, Anthony J.	12	M	2352
RUS		Agryskina, Nadezda Ivanovna	13	F	2538
RUS		Odrov, Viktor Anatolievich	13	M	2432
GER		Kuntze, Andreas	13	M	2397
ISL		Ásbjörnsson, Ásgeir Páll	13	M	2391
LAT		Kovalenko, Igor	14	M	2429
NED		Hooft, Diederic 't	17	M	2390
USA		Stein, Kurt W.	19	M	2533
IND		Sengupta, Deep	19	M	2515
GER		Lembeck, Karl-Heinz	19	M	2512
NED	CCE	Jong, Jan Willem de	21	M	2503
GER		Schmidt, Jörg (* 1954)	20	M	2376
POL		Gorzkiewicz, Łukasz	23	M	2473
DEN		Jensen, Tommy	23	M	2416
SRB	CCM	Petronijević, Zoran	23	M	2398
BEL		Van Assche, Jeroen	24	M	2393
AUT		Komaromi, Gabor	25	M	2401
ENG		Adair, James	26	M	2463
PAK		Idrees, Muhammad	25	M	2368
RUS	LGM	Churakova, Natalya Evgenyevna	28	F	2438
RUS		Bukarin, Mikhail Yurievich	29	M	2471
INA		Margana, Adhy	28	M	2359
ISL	CCM	Guðmundsson, Elvar	30	M	2504
NED		Werten, Tony	31	M	2392
ITA		Rombaldoni, Denis	33	M	2401
SUI		Schmid, Pablo	35	M	2525
IND		Dutta, Amit	34	M	2395
RUS		Kudryavtsev, Dmitry Aleksandrovich	34	M	2385
POR	CCE	Vasconcellos, Renato	35	M	2411
SLO		Coklin, Marko	37	M	2435
RUS	LGM	Matveeva, Maria Aleksandrovna	39	F	2498
IND		Das, Arghyadip	37	M	2363
CUB		Fernández Martínez, Juan Carlos	47	M	2393
ROU	LIM	Stanila, Elena	53	F	2389
USA		Kulick, Neil	59	M	2354
CRO	CCM	Feletar, Darko	79	M	2391
RUS		Matvienko, Vladimir Petrovich	18	M	2404
ENG		Rallabandi, Praveen Kumar	112	M	2398
ESP		Moreto Quintana, Alex	120	M	2437
UKR		Khanas, Valeriy	125	M	2357
SLO	CCM	Zajšek, Franc	147	M	2372
WLS	CCM	Yeo, Gareth	153	M	2378
UKR		Mashchenko, Vladimir	30	M	2481
SLO		Pokrivač, Izidor	32	M	2405
RUS		Rybin, Anton Sergeevich	24	M	2480
SWE		Qwarfort, Fredrik	26	M	2448
INA	CCE	Sitorus, Yosua	238	M	2403
CUB		Pérez Rodríguez, Rubén	38	M	2392
RUS		Mannanov, Rinat Rafikovich	56	M	2396
GER		Marwitz, Ullrich	78	M	2370
ITA	IM	Piccirilli, Fabrizio	74	M	2390
RUS		Enin, Anatoly Nikolaevich	107	M	2353
CAN	CCE	MacTilstra, Ian	160	M	2354
ENG	IM	Soh, Edmund	240	M	2403
WLS	IM	Claridge, John B.	255	M	2399
CAN		Czerniawski, Andrew	79	M	2436
GER		Ulbig, Stefan	107	M	2392
GER	CCM	Hartl, Hermann	124	M	2385
PER		Quiñones Borda, Jorge Victor	141	M	2406
USA	CCM	Koo, Oliver	150	M	2388
GER	CCE	Meißen, Frank	85	M	2401
CZE	CCM	Hrubčík, Martin	234	M	2412
GER	IM	Anderskewitz, Ralf	341	M	2401
UKR	IM	Koshmak, Iurii	109	M	2451
NOR		Johansen, Anders Sten	467	M	2386
USA	CCM	Hernandez, Angel	177	M	2401
AUS	CCM	Roebuck, Derek	110	M	2378
ITA		Galliano, Giovanni	113	M	2428
SWE	CCM	Strömberg, Håkan	121	M	2392
ENG	CCM	Carr, Trevor	229	M	2391
RUS	SIM	Panitevsky, Ivan Anatolevich	180	M	2504
HKG	IM	Tsang, Hon-ki	168	M	2419
SLO	CCM	Pirš, Jernej	185	M	2411
ESP	IM	Sánchez Huerga, Aser	193	M	2445
CZE	CCM	Binas, Jindřich (*1954)	237	M	2421
GER	CCM	Bär, Lutz	150	M	2400
GER	CCM	Lukas, Norbert	280	M	2355
BUL	CCM	Petkov, Stamat	126	M	2407
GER	IM	Wenzel, Stefan	177	M	2461
GER		Hesse, Olaf	293	M	2447
ARG		Fernández, Javier Horacio	329	M	2416
AUT	CCM	Hengl, Christian	236	M	2420
GER		Keskowski, Thilo	201	M	2406
GER	LGM	Achatz, Kirstin	223	F	2377
USA	CCE	Biedermann, Kyle	312	M	2384
GER	CCM	Tiemann, Christoph	162	M	2412
GER	CCM	Homont, René de	276	M	2353
WLS	LGM	Sherwood, Helen	452	F	2386
RUS		Razumikhin, Andrey Mikhailovich	364	M	2391
RUS	CCM	Novikov, Sergey Vasilievich	249	M	2400
RUS	CCM	Prozorovsky, Vyacheslav Grigorievich	208	M	2404
USA	CCM	Landes, Eric	222	M	2397
GER	CCE	Boos, Markus	193	M	2400
CUB	CCE	Marrero Rodríguez, Alexis	167	M	2416
ITA	IM	Mauro, Lucio	735	M	2407
AUS	CCM	Mulligan, Barrie	276	M	2414
CUB	CCM	Pérez López, Alberto	536	M	2407
ENG	CCM	Weldon, David J.	227	M	2410
POL		Mostowik, Daniel	773	M	2366
ITA	CCE	Bellegotti, Giorgio	243	M	2453
LTU		Voveris, Saulius	271	M	2388
ENG	CCM	Brasier, John	352	M	2428
USA		Zaas, Peter S.	224	M	2369
DEN	CCE	Konstantinov, Maxim	207	M	2421
ROU	CCE	Taras, Iulian	546	M	2390
RUS	CCM	Kozlov, Aleksandr Anatolievich	237	M	2382
POL	IM	Broniek, Mariusz Maciej	523	M	2439
USA	IM	Schakel, Corky	782	M	2416
ROU	CCE	Stanescu, Teodor-Adrian	146	M	2406
GER	CCM	Scheiba, Manfred	218	M	2403
CZE	CCE	Dědina, Miroslav	578	M	2354
GER	LGM	Bolz, Barbara	293	F	2402
GER	CCM	Zielasko, Andreas	284	M	2390
LTU		Voveris, Gediminas	660	M	2352
ESP	IM	De Carlos Arregui, Inigo	804	M	2433
ENG	CCM	Thompson, Brian	953	M	2413
ESP	CCM	Conde Poderoso, Antonio	597	M	2373
GER	IM	Löffler, Werner	802	M	2394
GER	IM	Langer, Raimund	443	M	2406
SUI	IM	Janisch, Manfred	658	M	2441
CZE	SIM	Sýkora, Josef	1477	M	2416
POL	IM	Mirkowski, Piotr	457	M	2412
GER	SIM	Gromotka, Harry	1456	M	2389
CZE	CCM	Cvak, Rudolf	1585	M	2363
ENG	SIM	Rawlings, Alan J. C.	1775	M	2393

 

 

 

 

 

Close examination of the players on the list will bear dividends for the ambitious player. What do these players have in common? All have come through the system and raised their ratings up to 2350/2400+. This group is of more interest to us, than, for example, players rated 2450+ , who often only play against their rating peers and draw,draw,draw!

Practically these players make up around 1% of the total CC player pool. To put into context their performance - over the last 4 years they are scoring around 70% with White and beter than 50% with Black.

 

There are a number of different strategies in play.

 

Now I could give a summary of the learning examination of these players but that would take all the fun out of it for you!  Some pointers to help in your analysis…..

 

• How many games are they playing/completing?
• What level are they competing at?
• What Openings and they playing/not playing?
• What is their win –loss –draw ratios with black and white?
• Are there any commonalities in their playing style?
• Do they take draws? Short or Long?
• Do you know any of them?

 

Putting some effort in here will allow you to reap rich rewards as it will give you an idea of “the way”, and Where’s there’s a way, there’s a will!

 

 

Russell Sherwood  Wednesday, October 25, 2017

In recent years there has been a massive increase in the strength of Chess engines and their influence on Correspondence Chess has been significant, leading a number of critics to observe that CC is Computer Chess, not Correspondence Chess.

 

It Is true that a CC player playing without any computer support, at the higher levels, is putting themselves at a near fatal disadvantage but there is hope for the player wishing to play a human led game at this level.

 

How can this be? Much comes down to two issues:

 

• Looking at the nature of the how the engine works and its strengths and weaknesses. It’s widely accepted that even the best engines (not necessarily the highest rated) only have the Chess knowledge of a 2200-2300 rated player. The reason they have ratings of (depending on which list you look at!) 3300-3500 is due to their tactical prowess, which is related to their ability to review millions of moves in time it takes a typical human to look at a mere handful.
• The rise of the “engine jockey”: Players will powerful hardware/software and very little engine knowledge

 

 

So how can we, the poor human, make progress here?

 

The answer is twofold – utilising another engine (even in theory a rather weaker one) to check our moves for tactical holes – a form of error checking and attempting to exploit the engine's lack of knowledge. Now some more astute readers may be thinking that some Engines (Komodo for example) are rather better at positional play than others but whilst the better engines will know, for example, to occupy the open file , they are doing this driven by fairly simple formula’s rather than understanding the implication of the move which may be in 20-30-40 moves time.

 

How do we look to put this into practice?

 

• Openings. In general engines struggle in openings, left to their own devices, as unless there is a clear way forward the number of options becomes too large a number of variations to be calculated, occasionally leading the engine astray. Looking through CC games is becomes obvious where a player thinks deviates very early in the game, knowing the typical engine will give a poor response. Now as a human to make this work, you need to understand the opening in question really well, especially its strategic aims.  This approach was (and is) used successfully in the King’s Indian Defence, where most engines are fairly clueless, although it must be said that the effectiveness of this approach is diminishing as engine knowledge is increasing in this area.
• Deep Strategic middle game ideas. Another tactic used by “Engine Killers” is to steer the game towards tactically quiet, but strategically rich middlegames. This tends to be seen by players who utilise the Nf3/g3 type systems. Due to the flexibility of these systems opening knowledge (books) often run out quite quickly and the evaluation of the engine either heads towards 0.0 or is stuck on some random number (this is caused by some positional evaluation factor of the engine which may have no relevance in the position). If in this position you have a clear plan with quite distant objectives, which the engine is looking for tactical and basic positional style moves you opportunities become clear.
• Engines can, sometimes, fall to a method related to the “slippery slope”. Imagine playing a game, materially is even, but your opponent has some significant positional plusses and you are under pressure. If you could free yourself from the pressure by giving away a pawn, in a non-fatal manner, you probably would, especially if it gave you some opportunities of your own. The engine is unlikely to do this. In this situation, it probably has an evaluation of about -0.4, but to give up the pawn would change this to -0.8 whereas to continue to play in the significantly pressurised would only lead to -0.5 or -0.6. This seems sensible but what then happens is that the engine goes down a slippery slope and 10 or 15 moves later is looking at an evaluation of -1.5 and a lost position!!
• Endgames have always been a struggle for engines. Tablebases have “covered this up” somewhat but it is not uncommon to see massive misevaluations. Now as a human we can look to play against this directly or in a much deeper way in the middlegame. If we know that in this type of position a win is much more likely with a b pawn rather than a pawn and if his bishop is not on the board a win is  even more likely ,then  our play 20 or 30 moves earlier can steer the game in this direction, way beyond any meaningful horizon the engine may have!

 

So my dear human’s all is not lost and it is possible to make progress against the metal (and digital ) monsters and their poor slaves but only if we (a) Utilise where we are stronger than the engines and (b) use the engines to nullify their own strengths.

 

Should you be successful, don’t look to repeat your success in the exact same line as, like the Borg in Star Trek, your opponents will quickly adapt!

 

As a final thought – there were some games with GM’s + obsolete engines vs Leading edge engines, where the Leading Edge engines prevailed. This is something of a red herring for us as whilst the GM’s are almost certainly much better players than us, their approach was to play in their own style, rather than look to exploit the engines weakness. In addition, the time control’s for this was for OTB, which changes the nature of the game somewhat!

 

Good Luck!

 

 

Russell Sherwood  Thursday, August 17, 2017

Have you ever got yourself into major time trouble? If this happens, especially in a number of games at once it can be a major headache and for some a cause of stress.

 

So having got into this hole, how do you get out of it? Here are a few ideas:

 

1. If you have vacation time – use it, whilst you analyse your games. It still surprises me how few players don’t do this.
2. Know the time controls – what exactly do you have to do? If you are playing in a 10 moves in x Days event, then once you reach each 10^th move milestone you will get an injection of time. Remember to consider the so-called “Free Day” as well. This information is at the bottom of the Tournament Summary page.  It is worth noting that not all time controls are of this type – some more exotic experiments don’t have time injections.
3. Consider your position in each event to play – is it a team event, are you chasing a Norm?  This will aid you in your prioritisation
4. Consider the position at the board in each game? Where is it on the Win to Loss spectrum of results
5. You now need to consider your plan in each game:
   1. Team events should, in general, be prioritised.
   2. If the game is “lost”, then resign it – why waste time you don’t have in these games?
   3. If the game is drawn or drawing then consider offering the draw
6. Having followed step 5 we should have reduced our “pile” a bit for our remaining games we need to consider:
   1. Have I got to move daily?
   2. How much time do I have available (both in physical and processor time)?
   3. What is the game situation?
7. We now have the grunt work to try and get to the time control. A few things can help:
   1. If the game has a fairly obvious line – go down it, using conditional moves if they are available in the event.
   2. If your response is obvious in a position – make the move (subject to a quick sanity check)

 

Hopefully after a few days of doing this you will have got yourself out of the immediate hole. Now is the time to consider how and why you got there in the first place.

 

I personally follow the 10 day rule – I rarely allow games to go below 10 days on my clock or not make a move for more than 10 days in a game. I probably manage to follow this 95% of the time and it prevents me getting into time trouble too often!

Russell Sherwood  Sunday, July 30, 2017

The Universal Solder?

If you spend any time analyzing with engines and have any ability yourself It becomes obvious that whilst they play close to perfect chess in some positions, they are also clueless in other positions. A traditional example being the King’s Indian Defense, where most engines struggle. An added dimension to this is that different engines excel and misbehave in different types of openings (and/or positions).

 

Knowing which ones work well (or otherwise) where is a significant potential advantage to the player. Trawling some of the Forums and Webpages can lead to a partial picture, although the problem is that any information gained this way is suspect and could become obsolete in the next version of an engine! If you are wondering why this is so? In simple terms most (stronger) engines are very finely tuned. This tuning means that the engine will play the “best” move in the highest number of positions. An improved version will means the engines plays slightly more positions correctly. The flaw is that, whilst Version 2 might play 94% of positions correctly compared to Version 1 which played 92% of positions in the right manner it does mean core positions – it could well be that 6% of positions more are correct but 4% more are now incorrect. This is general is not an issue, unless, of course, the positions (or more typically themes in positions) are the ones which are important in your opening repertoire.  Of course an interesting conclusion from this is that it would be possible to tune an engine for a specific kind of opening but it would probably be weaker in general!

 

So how can do this ourselves with some degree of certainty? An approach which can deliver dividends is as follows:

 

1. Examine your own opening repertoire
2. Identify Key positions in these openings. This can be done from personal knowledge or Opening Books (both Electronic and physical)
3. Create an EPD file with these positions and the “Best” moves
4. Run a “Beauty contest” with a number of engines to determine which best “understands” these type of positions.
5. Use the winner(s) s part of your evaluation strategy in this opening.

 

Now some may wonder why I believe using the EPD approach is the way forward, rather than Engines Tournaments (either between engines or self-play). The simple reason(s) are : (1) EPD testing is faster and (2) Engine v Engine tournaments can suffer from the results being skewed by either the difference between the two engines or the oddities of an individual engine.

 

What is worth saying that Engine v Engine can be useful if you can determine which Engine your opponent uses as their main support and openings/lines could then be selected to utilize this information!

Russell Sherwood  Sunday, July 30, 2017

As I have mentioned a number of times most Chess Engines are not well suited to Correspondence Chess.  In fact many, if not most, Engine Analysis techniques methods are designed to cover some of the gaps left in engine design.

Many Engines can be modified for CC in their Parameter file and some of the sensible choices in this area were covered in a previous article.  There are a number of current engines that do significantly better than most in terms of CC performance. So (in no particular order)

Houdini 5 (and previous versions to a certain extent) – Tactical mode allows a much shallower, deeper search to be applied.

Matefish – a variant which delivers a similar functionality to Houdini’s tactical mode.

Komodo – probably the most configurable engine, “off the shelf”

CorChess  - A Stockfish variant with a number of developments related mainly to a more effective search method

Thinksfish – Another Stockfish variant, with a massive number of adjustable parameters and a special “correspondence mode”

CIFish – a combination of CFish (faster Stockfish variant) and CorChess

All of these are a big step in the right direction for CC players and well worth a place in your engine stable………

There are a number of other improvements which are possible but we will talk about the “Blueprint” another time!