Analinp glitch? : LUSENET : MARP Editors : One Thread

So, I was browsing through scores at random, wiping out some obvious slow-mos by Cronos, when I decided I'd check the .inps beating me on Zektor. (I was mainly curious because the high score was made on mame33 and still hasn't been beaten)

So I load up trusty analinp, and sic it on the top 3 scores for the game. Right away, JSW's firing rate shows up close to the 1 mark. I think to myself that that's strange, as JSW is usually a pretty clean player. Then I check Canuck's--he's averaging right around 2. The plot thickens.

Then I get to my .inp, and analinp politely informs me that the .inp cannot be checked, as the game has a strange frame rate ~40.

This creates a bit of a concern for me, and I don't really think I need to explain why to either skito or BenJos. Could one of you look into it please?

Thanx, Q.T.Quazar

-- Anonymous, September 03, 2000


the games frame rate is less than 60, it does not detect it in the mame33 games because the game name is not provided in the inp, but since the new games (like your recordings) are rpovided in the inp, analinp can do a look up of a framerate table and see that this game is not very usefull in determining slowdowns because it has a slow framerate. make sense? I think there is or we were thinking about putting a command line option of the game name, so that when trying out older inps that don't put the gamename in them, it could still be warned.

analinp -n zektor mame33zektor.inp

etc... i don't think we did that yet though.

-- Anonymous, September 03, 2000

Ah, yes, this reminds me of something. Something I have mentioned before, but have to admit I did not always remember myself during my sessions of disqualifying dozens of recordings. Framerate is very important. Not only for the original Analinp part (i.e. the autofire detection), but also for the InpHisto part in it. It is very hard to catch slowdown in games that run at 30 fps. The whole point of the decisions we make based on InpHisto output relies on what we know of how fast fast pressers can be on 60 fps games. People are, in a way, at least, very constant. It is very hard to force yourself to play a game at 50% speed and hold down the buttons and joystick directions twice as long. (My post on MvdV's Pnickies is a very good example.) People are very likely to still hold down the buttons for the same amount of time across games, and regardless of how fast the game runs. I have actually got enough experience now to be able to recognise people merely by their InpHisto output. Or to know when they run too slow.

There are some things that require us to be careful, though, when going mainly by the InpHisto output.

(1) FPS. Several games run at 53 fps. That means that someone who usually averages 4 frames in InpHisto, is likely to average 3.53 on a 53 fps game. If this is not taken into account, then this may result in recordings being disqualified that otherwise would have been given the benefit of the doubt. I was reminded of this by my own MK3 recordings. I made several that I didn't upload because it seemed to me that I was running the game below regulation speed. I finally decided, I didn't. MK3 runs at 53 fps, and I am a "panic player". I don't hit a button once, when I'm in danger, I hit it lots of times. Quickly and in bursts. I see some analogies with T&F here.

I can say with certainty that my last MK3 upload ran at a constant 100% speed on my machine. Hell, it even runs at 100% with sound ON, but I played it without sound. I can record MK3 over and over again, using no sound and ridiculous frameskips, but the statistics will remain the same. If someone doesn't trust my recording nevertheless, I'll delete it, but there is no way I can make my statistics look "better" other than running the game at a speed that is too high.

(2) Some games have "strange" things in their drivers. Gaplus during T4 showed that. For some reason, Gaplus puts a lot (over 50%) of length 1 states in the recording. If you are not aware of this, you might disqualify a clean recording. That's why I usually compare .inps to other .inps on the same game (preferably done on the same version of MAME). If all of them show similar strange things, I'll record one myself.

Most vector games run at 40 fps. If you'd accept an average of 4 by someone on a 60 fps game, then you should accept a 2.67 on a vector game...

Cheers, Ben Jos.

-- Anonymous, September 03, 2000

Moderation questions? read the FAQ