Does +3 mean it's possible to convert a win without your opponent making a mistake or can it also be a draw and the +3 is just because the position is better even though it's not winning?
Using the analysis board after the game and using the top recommended moves I ran into two positions that seemed to be draws but the engine said around +3. One with my queen on c1 and king trapped on a-b files and my opponent with his pawn on f2 (starting position not about to queen), king on e6 and queen delivering checks. Another with my opponent left with pawn on a2 (again starting position), king in that corner, my pawn on a3 and I have a dark square bishop and king.
It can also be a fortress being a piece up. Like Harry & the wrong Bishop Thingy.
Stockfish evaluate the fortress, so the analys should reflect this.
I think I have never seen draw for evaluation > 1,5.
Please show as the position of possible.
There are still some types of position that engines will evaluate higher than +3 without there being a forced win. I put the latter situation you described to an engine on "self-play" mode. Evaluation was well over +3 throughout and then the 50 move rule kicked in and draw agreed. The program sees the benefit of having an extra piece and well advanced pawn but does not appreciate that the advantage stays the same for move after move.
On trying the position with Stockfish, it evaluated in similar fashion but when allowed to progress through the analysis it eventually reduced the advantage to a negligible level.
Yes, Stockfish cannot properly evaluate fortresses such as the one in your second case.
Basically, it has to analyze to depth 100 (50 moves) to realize that ALL possible variations end in a draw by the 50-moves-rule. Obviously, this takes some time, and in the meantime you will see that +3 score.
In your first case however, I wouldn't be so sure whether the game truly is a draw. Perpetual checks are much easier to calculate in comparison (if your king is trapped on 2 files, that greatly reduces permutations of possible checks), Stockfish will usually be able to identify them. Can you show the position?
Just try to win! :D
Move 64 we both promote to Queens, he plays pawn to f4 on move 73 which quickly leads to a draw but playing the recommended moves on the analysis board seemed to be a draw.
Actually it is a tablebase win. If you click on the little "book" symbol in the analysis you'll see "distance to mate" and stuff like that,
+3 doesn't mean anything. In contrast, mate scores are not falsifiable (actually mean something):
You can't post in the forums yet. Play some games!