Blitz vs Long Time control -chess engine functionality

[Darkmoon]

For Chess engine strength I've been lead to believe Long Time Control games are one of the best determining factors in getting a "sense" of comparative chess engine strength-not definitive-just a sense of their position in strength. Possibly a 100 game match at 2hr/40 repeated?!

Whereas running Blitz games would require quantifiable amount of games using a specialized gui.

My question is this -running blitz matches, with contrivances in opening books, et al- followed by predictions of which engine is superior is ...rather hasty.

But more to the point and this is my question!

Are the programming of chess engines like Komodo and Stockfish with their fast search and advanced eval function, making comparisons between them in blitz more difficult; when -if I am correct- these chess engines need longer time controls to function at peak capacity?!

Has their programming made them ineffectual to an extent at blitz time controls?

Please be kind in your response -I'm not a programmer. I haven't a clue if this is a dumb question or not.

Are these chess engines at a disadvantage when played at Blitz?

[jdart]

More search time improves play. However, modern engines are very strong at blitz. Also hardware keeps advancing, so blitz games today on modern hardware reach search depths that engines struggled to reach in long time control games a few years ago.

I think the only reason engines may be handicapped at blitz has to do with time management. Obviously you don't want it to lose on time, but also, time management may have to be adjusted to play optimally at blitz. I suspect my engine has some issues in this area because its relative rank is a bit higher at standard time controls than blitz. Still, it is not a big difference.

--Jon

[Eelco de Groot]

For Chess engine strength I've been lead to believe Long Time Control games are one of the best determining factors in getting a "sense" of comparative chess engine strength-not definitive-just a sense of their position in strength. Possibly a 100 game match at 2hr/40 repeated?!

Whereas running Blitz games would require quantifiable amount of games using a specialized gui.

My question is this -running blitz matches, with contrivances in opening books, et al- followed by predictions of which engine is superior is ...rather hasty.

But more to the point and this is my question!

Are the programming of chess engines like Komodo and Stockfish with their fast search and advanced eval function, making comparisons between them in blitz more difficult; when -if I am correct- these chess engines need longer time controls to function at peak capacity?!

No, not usually. Because the only way to test and develop the top engines, is to play literally 100 of thousands of games. There is no other way to get statistically relevant results. 100 games is of no use to test a single change and you just have to test every single change.

A side effect of this is that the top engines have been tested over and over again at very fast time controls, but long time control testing is done only by a few enthousiasts and by the rating lists, CEGT, CCRL, SPCC, and a few others. It is almost inevitable that some of the changes do not work the very best at very long time controls. But because there are many changes, it is still highly probable that the net effect is still good. I believe Don Dailey thought that Stockfish exhibited some signs of being developed and tested a bit in the direction of Blitz, he posted that in this forum.

Has their programming made them ineffectual to an extent at blitz time controls?

Please be kind in your response -I'm not a programmer. I haven't a clue if this is a dumb question or not.

Are these chess engines at a disadvantage when played at Blitz?

[Darkmoon]

For Chess engine strength I've been lead to believe Long Time Control games are one of the best determining factors in getting a "sense" of comparative chess engine strength-not definitive-just a sense of their position in strength. Possibly a 100 game match at 2hr/40 repeated?!

Whereas running Blitz games would require quantifiable amount of games using a specialized gui.

My question is this -running blitz matches, with contrivances in opening books, et al- followed by predictions of which engine is superior is ...rather hasty.

But more to the point and this is my question!

Are the programming of chess engines like Komodo and Stockfish with their fast search and advanced eval function, making comparisons between them in blitz more difficult; when -if I am correct- these chess engines need longer time controls to function at peak capacity?!

No, not usually. Because the only way to test and develop the top engines, is to play literally 100 of thousands of games. There is no other way to get statistically relevant results. 100 games is of no use to test a single change and you just have to test every single change.

A side effect of this is that the top engines have been tested over and over again at very fast time controls, but long time control testing is done only by a few enthousiasts and by the rating lists, CEGT, CCRL, SPCC, and a few others. It is almost inevitable that some of the changes do not work the very best at very long time controls. But because there are many changes, it is still highly probable that the net effect is still good. I believe Don Dailey thought that Stockfish exhibited some signs of being developed and tested a bit in the direction of Blitz, he posted that in this forum.

Has their programming made them ineffectual to an extent at blitz time controls?

Please be kind in your response -I'm not a programmer. I haven't a clue if this is a dumb question or not.

Are these chess engines at a disadvantage when played at Blitz?

As I referenced above I am aware of the quantifiable Blitz games necessary to show how well the chess engine is functioning or not.

Something I suspect that most "enthusiasts" are aware of "BUT" refuse reconciling themselves to for the sake of immediate self-gratification in posting their results. At least that is my impression.

But beyond that-if you don't have the numbers and you insist upon posting a few Blitz games -my first thought was how by comparison to long time controlled games-that Blitz games were by no means an accurate a tool for measuring chess engine functionality. That was my thinking.

But now my question has changed!

My assumption now is that Blitz games give totally different info than Long time controlled games.

I am not talking quantifiable levels of Blitz games! -I've rarely seen that on any forum -rarely with thousands or even hundreds of presented Blitz games -its happens but it is not the rule.

Long time controlled in the hundreds-yes!

What are Blitz games good for as indicators of chess engine functionality when given limited quantity-very limited-as is normally the case!

And like wise the same for long time controlled games.

[Eelco de Groot]

Well, then I think it comes down to the fact that nobody has the numbers. If you are not willing to play at least a few thousand games at long time control, then a claim that the result will be different than at the usual testing time control is just a claim. Most you can do is point to a trend if you gradually increase testing time. Larry K. is fond of doing that

But on the other hand, I do believe there is still considerable room for progress, so why should it not be possible that an engine X exists that can beat the living daylights out of both Stockfish and Komodo and it is possible that such an engine would perform better at long time controls. On the other hand at long time controls, the draw percentage will be higher, no matter what the difference in strength between two engines. And maybe I'm mistaken but then I think the scoring percentage of the stronger engine will go down, even if it plays almost perfect chess. I'm not sure of the math but at some point I think the weaker engine will not be losing every time, if the time control is long enough. The weaker engine will still lose some games of course.

Looking at the quality of play if you just have a handful of games, will be more convincing than the win-draw-loss statistics. But it is not easy to measure the quality of play with a statistical tool.