winMBT low percentile hits

Weasel · #1 July 5th, 2005, 04:06 PM

Glad I am not the only one, this feature of the game really annoys me.

Shadowcougar · #2 July 5th, 2005, 05:06 PM

I haven't noticed that. I find the hits in a game seem to even out the % of hit/miss.

OF course I have had the battle when I had a Cav tank plt fire at a on coming T-62 12 times and miss all of those shots. also had a hit at 4000 meters also.

kevin · #3 July 5th, 2005, 05:14 PM

I thought that I had once noticed this, but concluded it was a psychological thing. I think we are more likely to remember a 55% miss than a hit, as well as remembering a 20% hit more than a miss, because those things go against our expectations.

I guess the only way to double check is to get a mechanical counter and do an experiment.

Uncle_Joe · #4 July 6th, 2005, 03:16 AM

I also see most 40-50% chances as a near auto-hit. I cant understand how 5 out of 6 shots that supposedly have a 40-ish percent chance to hit are scoring.

I see this quite consistantly too. Like the rest of you, I figured it was just me too.

Starmyth · #5 July 15th, 2005, 04:20 PM

Is Kevin really offering to do a statistical analysis for us?

PlasmaKrab · #6 July 15th, 2005, 04:33 PM

You can do one yourself, should be no big deal, only some work...

Create a test unit with lotsa shots, quite uneffective weapon, note the accuracy value, unit viz, stabilisator, .
Create a map on which you put several of these with standardized M/E ratings, face them with many heavily-armored unarmed target units, which you place at different ranges.

Then have a bout of firing and note for each shot the given PK and if it hit or not.

When you have the data, sort by probability and look at what you get. Again and again

Yeah, that is tiresome, and I guess you can get a automated way to do it...

Anyway I don't think you will come up with anything surprising, but if you have 50 hours to spend...

Starmyth · #7 July 15th, 2005, 04:40 PM

Maybe someday. If I get really really bored.

loktarr · #8 July 15th, 2005, 05:45 PM

In order to avoid you some boring and long experiments, I'll give some useful informations:
1) Given percentiles are main weapon ones -> try to kill a drone with T-62 main gun, 1-3% seems like elite gunner in your T-62, with machingun it's pretty better....
2)The given percentile depends only on units. Explanation: take two M1A1 one with a 20 experience the other with 120. at the same distance of each deploy a T-62. Given percentile should be the same... Try the fire and you will see. Experience is only tested after you decided to fire at the unit, not before...
if that's true, it will explain everything, but it may have changed between SPMBT 3.01 and WinMBT...

kevin · #9 July 16th, 2005, 01:00 AM

Starmyth,

noooooooo. sorry I don't have the time for that. As usual my friends loktarr and plasmakrab seem to be 'over engineering' a test.

You don't need to set up special scenarios.

Set up an excel sheet, or a scratch piece of paper. Set up 100 columns and then divide each of those into a fired / hit column (total of 200.) The next time you play SP, keep track of the estimated % and place a tick mark mark in the correct fired column. If it hits, put a mark in the hit column. After 100 fired ticks, you ought to have an idea if SP is estimating coreect hit %'s.

PlasmaKrab · #10 July 16th, 2005, 04:59 AM

Sorry Kevin, I wasn't meaning that you were doing it, I don't even knkow where Starmyth found the info, so Iwasn't sure you wouldn't either...

Either case I was just indicating that he could do it himself, same way as you explained, only with better standardisation... Not necessarily useful though.

But if you want to have statistically useful results, you ought to allow for many shots for each PK %. One each won't do.