A few days ago, I reviewed several Chess related OCR scanners that convert hand written chess score sheets used in chess tournaments into pgn files that can be used by chess programs and engines for review. Here was the last article, if you missed it: https://alaskalinuxuser3.ddns.net/2024/07/12/a-quest-to-find-the-best-ocr-to-convert-handwritten-score-sheets-to-digital-pgn-files/
One of the apps I reviewed was “Chess Digitizer”. It is a mobile app which was pretty handy when you consider that with your cell phone, you could review your games between rounds at a tournament, or do the uploads “on the go”. I am not really in the app review business, but as the developer reached out to me after my last review, I felt it was fair and appropriate to review it again with the latest revision. The developer read my review and pointed out that the version of the app I was using was several months old, and some of my critique’s have been resolved. He asked me to consider trying his new app and giving another honest review. I will note that the app is subscription based, with some parts of the app being available for free, and some being available to the pro users who have a subscription. I decided to test it both ways and see how well it worked. Full disclosure, the developer gave me access to the pro/subscription version for free to test the app. The app is now called Chess Scanner.
In any review, there is a subjective part, based on how we feel, and an objective part, based on facts. I plan to use both of these here, allowing the reader to draw the best conclusion. However, for the objective test, I am not sure how fair it is to use the same score sheets again, as in “would my using and fixing them last time alter the algorithm specific to these samples?” That I don’t know. To be safe, I did re-run two of the old score sheets, and added two newer sheets.
Subjective review:
Just as I mentioned in the last post, the GUI for this app is fairly good and intuitive. I will note that the color scheme for the latest release got even better, and it looks like a slick product. I like that it shows you the scoresheet while you are working through the moves, I also really like how easy it is to tap and change the moves. The layout is good and easy to use while holding your phone in your hand.
There are very few things about the app that I am not thrilled about. But as it is a review, I’ll list them here. First, there is a huge gap in usefulness between the free and pro versions. Essentially the free version is a trial version to see if you like the app and if it will work with your hand writing. While both will read your hand writing and make a digital pgn, you can only download that pgn, or send it to Lichess, Chess.com, etc., with the pro version. If you have the free version, you can “share” the game by sending someone a link to the Chess Scan servers, where they can view the game, but still don’t have access to the pgn. I completely understand, they need a way to financially support the hardware and software development, but you will have to pay for the services if you want a tangible pgn that you can copy/paste into any engine.
I do want to clarify here, all of these applications that I have tried cost some sort of fee to use, which is appropriate, since they do need to cover the cost of running their application in the cloud or on their servers. I am not sure what is the best method for charging the customer, though. Chess Scanner works on a subscription, once subscribed, it appears you can upload as many sheets as you want. Other applications, such as pgnApp are based on purchasing a number of games to upload. For me as a player, I only play in one to four tournaments a year, so I am not sure a subscription would work well for me. However, as a tournament director, I have access to 90+ score sheets from a single tournament. So that would be the better deal. Just depends on your needs.
Secondly, I got rather used to seeing the board when editing the score sheets in pgnApp and PawnParse. The screen being smaller on a tablet or phone, I completely understand their design choice to show you the score sheet and the editing tools, rather than the board. But sometimes, when a score sheet is not clear, it helps a simpleton like myself to be able to see the board at this point in the game, so I can determine if Re7 was Ree7 or Rfe7. Often by looking at the board and the subsequent moves, it becomes readily clear. Without the board, I have had to guess a few times and then go back to change my guess, since that was not the correct guess. I would recommend that the developer consider having a toggle switch or button so you can somehow view the board in it’s current position, this would really help with clarity. This is principally a problem when the score sheet is incorrect, which seems to be 75% of the time when I upload other people’s scoresheets. (I’m not perfect, but for some reason I can legibly write the correct moves of the games I have played.)
While great strides have been made in the score sheet alignment department, I still had one of the four score sheets I tested with that did not line up well and required several pictures until it got it right. As I will discuss below in the objective review, I had more trouble with the tablet taking the picture than with the phone.
And finally, for my regular readers, who typically do not use Google services, you do have to have Google play services on your phone/tablet to use the pro version. Seems odd, but that is something I tend to avoid, so to use the pro version I had to borrow a tablet from my wife to test it. However, that is only a problem for a niche group of people.
Objective review:
For the objective portion of the review, I fed it several game sheets and just wrote down the number of half moves and the number of moves it guessed incorrectly. In my testing, I found no major difference (actually, no difference noted at all) between the basic (free) and pro (subscription) version of “Game Generation”, where you choose how you want it to recognize the hand writing if using a legible score sheet written by someone with good hand writing. However, it did seem to help when using the pro version for someone who had terrible hand writing.
Note that the accuracy percentage I am giving to these moves/score sheets is not the accuracy rating the app gives itself during the recognition process. My percentage is based on the total number that were correct after refactoring for each fixed move.
On either the basic or pro version, when I fed it a very legible score sheet, written in my own hand at one of my games, it recognized all the characters and gave me a proposed pgn with only 1 move that was in error of what I had written. That means it correctly interpreted 50 out of 51 half moves, or 98% accuracy. I then fed it another of my hand written score sheets where it correctly interpreted 99 out of 100 half moves, or 99% accuracy.
That said, I fed it the dreaded sample sheet #2 from the previous post, which before it could not read at all (the hand writing is good, but the score sheet is a half sheet/small sized sheet. This time my results varied. With my tablet, Chess Scanner was only able to read the last column of moves, no matter how many times I took the picture. In a game of 41 moves, it could only read the last 11, none of which were legal without the first 30, so it did not do well. Then I tried the same score sheet with my cell phone, and it was able to read the score sheet with 12 errors for 82 half moves, or 85% accuracy. Again, the hand writing was perfectly legible, just the scoresheet had small boxes for the moves. Is this an app problem or a hardware camera problem? I’m not really sure.
I would like to note, though, that it also read the bottom of the page as part of the moves. (Please see the picture.)
I then fed it the challenge score sheet #3 from the previous post, which has 103 half moves, 6 known errors, and very bad hand writing. Using the pro version, it got an impressive 80% of the moves correct, which was objectively better than the competitors last time, having only 19 failures of recognizing the moves or characters. I don’t have access to the software, but I think a big factor in determination seems to be forcing legal moves. If the moves of the game were legal, then it does pretty good at figuring out what move must be next.
Conclusion:
Overall, I am really glad that the developer reached out to me for a second look at this app. It had been updated and improved since the version I tested before, and it is a really great app. The interface is really good, and it did exceptionally well with good hand writing, and really good with poor hand writing. I could see someone using this between rounds at a tournament to see what went well/wrong in their last game, and making it quick to share your games online with friends. This app seems best suited for the player, especially an active one. As a tournament director, I would probably rather crunch 20 to 90 score sheets on a computer with a mouse and keyboard, but this is a really handy tool to have in your pocket.
Linux – Keep it simple.