I think it would be pretty cool to be able to determine a winning percentage based on the score at any given time in a match. For instance, with 5 minutes left in a Team Deathmatch, Team 1 has a score of 4500 and Team 2 has a score of 4000, through the sample's I will be able to determine a winning percentage based on that (and all times and scores). Say, that through the samples a team with a 500 point lead (Team 1) with 5 minutes left to go in a match, they have a 55% chance of actually winning the game.
Example 2: Team 1 has a 1000 (6000 to 5000) point lead with :30 left to go. Their winning percentage is probably much more higher, maybe around 95% (I'm just assuming, I don't have all the data I need) (with nukes and other last minute killstreak rewards etc... accounting for the 5% Team 2 could win.)
I finally got through my first game of data (it's a little more time consuming than I thought it would be). The first graph is the score of both teams throughout the game. The second graph is the graph of the difference between the scores.
If we were to extrapolate from this game only, we could conclude that if you get first blood, you have a 100% winning probability. Also, if you have a 1400 point lead after about 4 minutes played, you have a 100% winning probability. However, because this is only one game, we need more samples and do some averaging out at different times. As you can tell this was a pretty lopsided game until about 5 minutes into the match when Team 2 goes on a long streak and closes the gap between themselves and the other team.
This is just the first of many games I'm going to be watching over the next few months. Hopefully, I can get a fairly accurate model.
I'm still thinking about all this stuff and how and what I'm going to need. If any of you have any suggestions, let me know in the comment section.
Posts
I know i've posted on the mw2 forum but ill do it here too,
ReplyDeleteThink this kind of analysing would be more useful in games with objectives, such as domination, search and destroy, sabotage, as these are the games that much less varying that tdm or ffa where the objectives are more about the team as oppose to the lonely wolf of tdm
Domination I could definitely use a similar type of analysis and using the score of holding onto the flags instead of the kills for tdm.
ReplyDeleteSearch and Destroy and Sabotage would be harder and I'm thinking of a way to do that but since the data gathering is very time consuming I'm sticking with TDM for now (not even doing FFA right now).