Originally posted by no1marauderCould you give a few example positions?
I don't know if it's been fixed in Fritz 11, but earlier versions were constantly giving large numerical advantages to a side with a pawn up in an opposite colored bishop ending that any experienced player could tell was a dead draw.
I have Fritz9 and have never seen that occur.
The post that was quoted here has been removedNot exactly opposite coloured bishops 😛 But it's the same principle.
After 2 minutes it says +-4.55 but given the black side easily holds the draw.When you then go back it says +=0.33.Horizoneffect at work I guess.
Does Fritz8 hold the draw?Because it might evaluate a position wrong but at the same time play it right,making the wrong evaluation a moot point.
Originally posted by KatastroofThis is an effect of retaining hash tables. An engine often cannot see the true value of a move until several moves down the line but can then retain that knowledge if stepped backwards from the final position. This is why good interfaces do their analysis backwards (i.e. starting at the last move rather than the first) so that they have already assessed the positions that the good (or bad) move leads to.
Not exactly opposite coloured bishops 😛 But it's the same principle.
After 2 minutes it says +-4.55 but given the black side easily holds the draw.When you then go back it says +=0.33.Horizoneffect at work I guess.
Does Fritz8 hold the draw?Because it might evaluate a position wrong but at the same time play it right,making the wrong evaluation a moot point.
1 edit
Originally posted by BedlamYeah,I tried that of course.With a 1 pawn difference,not a few pawns.It always said =0.00,hence why I wanted to see some positions where it cannot spot the draw.Maybe it takes more pawns,like 3 vs 4 or 4 vs 5,will try that 🙂
Just drop a couple of oppsite colour bishops on the board give one side a few pawns and the other side none. Let fritz try and work it out.
- Joined
- 21 Apr 06
- Moves
- 4211
Originally posted by Katastroof3 vs 0 pawns in oppsite colour bishops is a draw a lot of the time, fritz will most likely get those wrong if you dont have tablebase.
Yeah,I tried that of course.With a 1 pawn difference,not a few pawns.It always said =0.00,hence why I wanted to see some positions where it cannot spot the draw.Maybe it takes more pawns,like 3 vs 4 or 4 vs 5,will try that 🙂
1 edit
Originally posted by BedlamYes,but 3 vs 0 would be much harder to calculate.I could understand it can't spot such a draw.But with just 1 pawn difference,as marauder stated,I'd be surprised.Would make Fritz look a little silly and that's always good 🙂
3 vs 0 pawns in oppsite colour bishops is a draw a lot of the time, fritz will most likely get those wrong if you dont have tablebase.
edit: and I don't have tablebases
- Joined
- 21 Apr 06
- Moves
- 4211
Originally posted by KatastroofYes but it will think even obvious drawn positions are won etc and will keep doing so for 40 or so moves while it pushes wood around 🙂
Yes,but 3 vs 0 would be much harder to calculate.I could understand it can't spot such a draw.But with just 1 pawn difference,as marauder stated,I'd be surprised.Would make Fritz look a little silly and that's always good 🙂
edit: and I don't have tablebases
3 edits
Originally posted by Katastroofthere's no draw, show it to me.
Does Fritz8 hold the draw? "Game 4873194" Because it might evaluate a position wrong but at the same time play it right,making the wrong evaluation a moot point.
Originally posted by no1marauder
I don't know if it's been fixed in Fritz 11, but earlier versions were constantly giving large numerical advantages to a side with a pawn up in an opposite colored bishop ending that any experienced player could tell was a dead draw.
I think most engines will say black is winning on this one and Berliner also has a very nice example on "The System" but since I am not at home I can't show it.