How To Convert Real Life Stats To Game Ratings

[ZoneBlitz]

I saw a couple of questions about how to convert real-life stats to ingame ratings so I thought I'd take a crack at explaining.

Actually it’s really easy to determine how to create a rating.

First, determine the range of the in-game rating. For example, you may want to use 25 – 99, 43 – 100, or 50 – 84, etc…

We then create two variables: RatingLow, RatingHigh.

Second, let’s assume we will assign the lowest actual statistic to the lowest rating, and the highest actual statistic to the highest possible rating. So, for example, I noticed on one NCAA stat site assists ranged from 1.2 to 6.7 per game (for all players with 20+ mins). We can then, using a 25 – 99 rating, decide that 1.2 = 25 and 6.7 = 99.

We now create two additional variables: StatLow, StatHigh.

Before discussing the methods, keep the following additional variables and terms in mind:

RatingRange: the total range between the highest and lowest rating (e.g. Awareness in the videogame)
StatRange: the total range between the highest and lowest statistic (e.g. Assists in real life)
RangePart: the ratio that relates each difference in the statistic to a difference in the rating
AttributeRating: the real-life statistic converted to the videogame rating
ActualStat: the real-life statistic for the player
“/” Symbol: divided by, for example; 6 / 2 = 3

Now, before we plug in the numbers, the method is as follows:

RatingRange = RatingHigh – RatingLow

StatRange = StatHigh – StatLow

RangePart = StatRange / RatingRange

AttributeRating = ((ActualStat – StatLow) / RangePart) + RantingLow

Now, we plug in the numbers. Keep in mind that I do perform some rounding of the numbers to keep it clear.

So, let’s say we want to convert a guard with 5.4 assists per game to an awareness rating with the in-game ranges mentioned earlier: 25 to 99, and the stat ranges also mentioned earlier: 1.2 to 6.7.

Using the method above, it’s quite simple:

RatingRange[74] = RatingHigh[99] – RatingLow[25]

StatRange[5.5] = StatHigh[6.7] – StatLow[1.2]

RangePart[.0743] = StatRange[5.5] / RatingRange[74]

AttributeRating[97.7] = ((ActualStat[5.4] – StatLow[1.2]) / RangePart[.0743]) + RatingLow[25]

Keep in mind that, you calculate the part in the parentheses first – starting from the innermost parentheses and working your way out from there to the final answer. So, in the final line of the method where you find “AttributeRating”, first subtract “StatLow” from “ActualStat”, then divide that answer by “RangePart”, and finally add the resulting answer to “RatingLow”. That will give you the “AttributeRating”.

If you have any questions or want to figure out how to reconcile multiple real-life stats into one in-game rating, just let me know. I’m a statistician and do this kind of stuff all day at work for my company.

[baylor_guero]

Re: How To Convert Real Life Stats To Game Ratings

I'm not sure if its that simple. You would also have to factor in context of the statistics obtained IMO. For instance, even if we find that X player has the most shot blocks per minute played, it might just be that X players just takes more attempts. In your example of assists, you would likely end up with the league leader in assists getting the highest passer rating, but he would not necessarily be the best passer. You would have to factor in turnovers, pace of game, type of offense being run, and quality of teammates.

All that to say, I think you can get a rating that approximates the actual quality of a particular characteristic for a player, but that falls short in that using only objective data, you haven't properly accounted for subjective information which has a material impact on the statistics produced.

[ZoneBlitz]

Re: How To Convert Real Life Stats To Game Ratings

I'm not sure if its that simple. You would also have to factor in context of the statistics obtained IMO. For instance, even if we find that X player has the most shot blocks per minute played, it might just be that X players just takes more attempts. In your example of assists, you would likely end up with the league leader in assists getting the highest passer rating, but he would not necessarily be the best passer. You would have to factor in turnovers, pace of game, type of offense being run, and quality of teammates.

All that to say, I think you can get a rating that approximates the actual quality of a particular characteristic for a player, but that falls short in that using only objective data, you haven't properly accounted for subjective information which has a material impact on the statistics produced.

I think you bring up some good points. Of course, you can factor in any context that you deem appropriate. Of course, the statistics I used were from those with 20+ minutes per game. Even in football, for example, the leading rusher may be in a run first offense, but that doesn't detract from him being a great rusher. Therefore, the league leader in assists is still a good passer - even if the offense is geared towards him being high in that category.

Returning to your point, any array of factors can be accounted for in a statistical model. As long as subjective data can be quantified, it can be included in a model.

Of course, for our purposes, I'm thinking that this simple approach can meet the needs of us as gamers who just want to have some fun :-)