I also just noticed that it has my location showing as the UK. I am in the US. 
Game Review info doesn't match Insights info
I also found a discrepancy in the game review from this screen to the final screen:
* note - this isn't an adjusted accuracy

One screen shows a mistake, and one shows an inaccuracy....
If you look at the only game I have in the insights for daily, its accuracy doesn't match the accuracy of the insights. I find this leads to a few problems in trying to determine how well we are actually playing or how well our opponents are actually playing. Is there a way that this could be made uniform, to avoid confusion and have a standard accuracy rating as it applies to both features?
If this isn't practical for now, which one should be considered more accurate? Is there a way to convert the inaccurate number to one of accuracy?
Insights is likely running at a different depth than what was used on your game. The process does all that separately from user ran reviews/reports. It also doesn't modify existing game analysis data, so if you ran and old report, insights doesn't update the game and previously unanalyzed games still don't have analysis.
Add to that, if that game was run on the old Game Report function (I haven't looked) then it's using a completely different analysis and move classifications algorithm.
I also just noticed that it has my location showing as the UK. I am in the US.
That's the location of your opponents.
I also found a discrepancy in the game review from this screen to the final screen:
* note - this isn't an adjusted accuracy
One screen shows a mistake, and one shows an inaccuracy....
That's the quick analysis, which is very low depth.
If you look at the only game I have in the insights for daily, its accuracy doesn't match the accuracy of the insights. I find this leads to a few problems in trying to determine how well we are actually playing or how well our opponents are actually playing. Is there a way that this could be made uniform, to avoid confusion and have a standard accuracy rating as it applies to both features?
If this isn't practical for now, which one should be considered more accurate? Is there a way to convert the inaccurate number to one of accuracy?
Insights is likely running at a different depth than what was used on your game. The process does all that separately from user ran reviews/reports. It also doesn't modify existing game analysis data, so if you ran and old report, insights doesn't update the game and previously unanalyzed games still don't have analysis.
Add to that, if that game was run on the old Game Report function (I haven't looked) then it's using a completely different analysis and move classifications algorithm.
If the game review is a newer feature, which I think is, it is ran on the newest feature.
I also just noticed that it has my location showing as the UK. I am in the US.
That's the location of your opponents.
Thanks for that. I was thinking OTB with that one for some reason from the get-go. ![]()
I also found a discrepancy in the game review from this screen to the final screen:
* note - this isn't an adjusted accuracy
One screen shows a mistake, and one shows an inaccuracy....
That's the quick analysis, which is very low depth.
I am pretty sure game review runs at 20 plies. DO you know the depth of insights?
That's the quick analysis, which is very low depth.
I am pretty sure game review runs at 20 plies. DO you know the depth of insights?
Game review, when you click the review button, runs at what you have it set to. I run most of mine at depth 26 and occasionally at depth 30. Though the quick post game analysis is at something like 8 or 10. Insights is 18 I believe, though I'm not 100% certain.
That's the quick analysis, which is very low depth.
I am pretty sure game review runs at 20 plies. DO you know the depth of insights?
Game review, when you click the review button, runs at what you have it set to. I run most of mine at depth 26 and occasionally at depth 30. Though the quick post game analysis is at something like 8 or 10. Insights is 18 I believe, though I'm not 100% certain.
I will set mine the same, so the data is closer to the same. Thanks for the info and I didn't know the review was adjustable.
@Martin_Stahl,
I checked all available depths and none of the numbers matched the insights data, using all of the game review options. It is either at a shallower depth or using a different engine.
@Martin_Stahl,
I checked all available depths and none of the numbers matched the insights data, using all of the game review options. It is either at a shallower depth or using a different engine.
Engines can have variance, so you can expect some differences. Though, as far as I'm aware, the same engine is being used for both.
@Martin_Stahl,
I checked all available depths and none of the numbers matched the insights data, using all of the game review options. It is either at a shallower depth or using a different engine.
Engines can have variance, so you can expect some differences. Though, as far as I'm aware, the same engine is being used for both.
They have a list you can choose from for separate analyses. I have seen some differences from one run to the next. It could even be a matter of what else I was using my computer for.
They have a list you can choose from for separate analyses. I have seen some differences from one run to the next. It could even be a matter of what else I was using my computer for.
Game Review is not impacted by the engine choices in settings. That's for position analysis.
My understanding is that everything is tuned to a single engine currently.
If you look at the only game I have in the insights for daily, its accuracy doesn't match the accuracy of the insights. I find this leads to a few problems in trying to determine how well we are actually playing or how well our opponents are actually playing. Is there a way that this could be made uniform, to avoid confusion and have a standard accuracy rating as it applies to both features?
If this isn't practical for now, which one should be considered more accurate? Is there a way to convert the inaccurate number to one of accuracy?