[ View Thread ] [ Post Response ] [ Return to Index ] [ Read Prev Msg ] [ Read Next Msg ]

BGonline.org Forums

transcriptions / A FRICKIN' GAMECHANGER ALREADY EXISTS!!!!

Posted By: sebalotek
Date: Thursday, 8 February 2018, at 7:51 p.m.

In Response To: transcriptions / A FRICKIN' GAMECHANGER ALREADY EXISTS!!!! (Casper Van der Tak)

What I was hoping to do is draw community attention to this software, which could significantly reduce the amount of human effort / the number of inaccuracies / the amount of confusion (and even arguments!) involved in the transcription of a batch of video recorded backgammon matches.

Before it gets shot down, I think we have to be somewhat realistic and manage expectations here..

IMO, even a skilled human transcriber will find it difficult to enter potentially hundreds of moves plus cube actions with 100% accuracy, first time, every time, every match. They may occasionally have to rewind sections, pause the video and even edit a few of their move inputs before they can finally run the analysis.

And in light of the 'bear selling' references, to make a fairer comparison I think we should not neglect to mention that even once a bear has been shot, there is still a great deal of preparatory work (cutting, cleansing, curing) to do before you can sell the skin (ask any hunter).

So I think it is not realistic to expect 100% accuracy from this project out of the box, even in its future 'finished' state. Variances in video clarity, resolution, dice colouring, board colouring, orientation, cocked dice occurrences, illegal moves and myriad other unpredictable factors will play a large part in its accuracy from random match video to random match video.

I guess a way it could be beneficial, even in its imperfect state, is...

  1. Chillax and run an automatic scan of the match video with Kisono's software
  2. Output a (txt? / mat? / sgf? / xg?) file which can then be imported by a standard bg bot
  3. Input verification (important!) - Once the data import is completed, open one window with the video file alongside another with GNU / XG.
  4. While the video plays, simultaneously step through each move with GNU / XG. Any visible conflicts should be fairly obvious. Manually correct any (hopefully infrequent) data source errors.
  5. Run your match analysis as normal :)

Of course, this workflow is just a suggestion and there may be even better ways. Multiple passes of step 1, followed up by some kind of averaging algorithm may increase accuracy.

I don't have figures, but even with 80% accuracy I believe this software could reduce the required human effort, time required and sheer drudgery of transcribing a match by (very rough guesstimate) 50% - an amount not to be sniffed at.

In conclusion, I think this is potentially one of the most significant developments in backgammon so far this century. IMO it is currently flying under the radar - and we would be lax to not proactively offer help and encouragement to MKisono in his endeavours.

BTW, has anybody actually tried to contact him about this yet? If you have, please keep us updated..

Messages In This Thread

 

Post Response
Your Name:
Your E-Mail Address:
Subject:
Message:

If necessary, enter your password below:

Password:
 

 

[ View Thread ] [ Post Response ] [ Return to Index ] [ Read Prev Msg ] [ Read Next Msg ]

BGonline.org Forums is maintained by Stick with WebBBS 5.12.