I noticed this because I happened to note the analysis values before.
For the line: 1. e4 e5 2. Nf3 b5 3. Bxb5 Bb4
At a depth of 25, SF 18 gave me an evaluation of +2.8. I can't run SF 18 anymore on Lichess but I can run SF 11 HCE and it says the same thing: to a depth of 26, the evaluation is +2.8. For the record, the highest I saw it go at any depth with SF 11 is +3.2. I tried boosting the settings as high as I could and running SF 11 for a while, and I got to a depth of 31, evaluation +3.2. I'm pretty sure SF 18 gave me similar results.
SF 19, on the other hand, gives it +4.0 at a depth of 25, and the lowest I saw it go at any depth was +3.7. I tried boosting the settings as high as I could and running SF 19 for a while, and I got to a depth of 29, evaluation +4.8.
So, what is going on? Is SF 19 just that much more accurate? Because SF 19 at the lowest depth I could see, depth 10, gives around +4.0. So is SF 19 running for like half a second to depth 10 more accurate than SF 18 running for many minutes to depth 30+?
I noticed this because I happened to note the analysis values before.
For the line: 1. e4 e5 2. Nf3 b5 3. Bxb5 Bb4
At a depth of 25, SF 18 gave me an evaluation of +2.8. I can't run SF 18 anymore on Lichess but I can run SF 11 HCE and it says the same thing: to a depth of 26, the evaluation is +2.8. For the record, the highest I saw it go at any depth with SF 11 is +3.2. I tried boosting the settings as high as I could and running SF 11 for a while, and I got to a depth of 31, evaluation +3.2. I'm pretty sure SF 18 gave me similar results.
SF 19, on the other hand, gives it +4.0 at a depth of 25, and the lowest I saw it go at any depth was +3.7. I tried boosting the settings as high as I could and running SF 19 for a while, and I got to a depth of 29, evaluation +4.8.
So, what is going on? Is SF 19 just that much more accurate? Because SF 19 at the lowest depth I could see, depth 10, gives around +4.0. So is SF 19 running for like half a second to depth 10 more accurate than SF 18 running for many minutes to depth 30+?
it might just faster. in reality there are only 3 outcomes of chess. So if see deep enough there is no difference between +16 and +1.
Also since stockfish had neural net evaluation then every generation will provide different values.
it might just faster. in reality there are only 3 outcomes of chess. So if see deep enough there is no difference between +16 and +1.
Also since stockfish had neural net evaluation then every generation will provide different values.