As a continuation from the SDA thread
https://forum.speeddemosarchive.com/post/grand_theft_auto_iii_replay_exploit_or_whatever.html
It has been improved by Paddy (or Patrick) and I want everyone's opinion/improvements to this trick.
Thanks, -Guy
So to summarize when it is beneficial to use this:
Both R(dur) and t(b) are of course averages for all the buffers you do. Using the time it takes to do a replay as 8 frames (per Patrick's suggestions it can go down to 1 frame when done optimally) and the time buffered per replay use (higher is better, but means more accuracy. This would be 0.5 if you hit the replay when the game is at 30 ingame seconds thus 500ms irl) as 0.8, this trick would be beneficial up to 45 seconds of buffering. Using R(dur) = 2 and t(b) = 0.8 this would go up to 55 seconds and a TAS would be able to make this beneficial up to 58 seconds (R(dur) = 1, t(b) = 0.97)!
You can use this link to calculate the maximum time where it still makes sense to use buffering with your own values.
Obviously you have to keep in mind that the more times you have to buffer, the more difficult it will be to find spots where you can do this without losing time (you can factor in any lost time by increasing R(dur) in the formula if you like).
So F1 creates, F2 saves, F3 loads, F1 cancels again. Do you then have to press F3 -> F1, F3->F1 every time or just F1->F1, F1->F1, etc.? (In order to lose as little real time as possible).
Another question: there was some talk of skipping some Maria dialogue. Can anyone explain where/how?
Thank you!