Total Value: A stat created to determine the MVP

Games Other Halo
1

When Halo: Reach came out, so did a new “rating” system. You start at 1000 and either increase or decrease based on your kills and deaths as the game progresses. It’s a step forward from the old, “most kills wins MVP” way, but I’ve thought up a way that incorporates just how valuable one is to their team. Total Value. Sure it’s possible someone thought of this before, but as you probably can see I have not been on these forums very long.

First, we’re going to determine the value you had towards your team’s score. This deals with Kills, Assists, and Your team’s final score.
((Kills + Assists) ÷ Your team’s final score ) +.1
*** I added .1 to prevent the impossible (dividing by zero)

Second, we’re going to find the value you had against your team’s score. This deals with Deaths, Penalties and Your opponent’s final score.
((Deaths + Penalties) ÷ Your opponents final score ) +.1
*** Again, dividing by zero is bad news.

The last part is quite simple. We’re going to take the value you had towards your team and divide that by the value you had against your team. This number is your Total Value.
(((Kills + Assists) ÷ Your team’s final score ) +.1) ÷ (((Deaths + Penalties) ÷ Your opponents final score ) +.1)

Case study #1: http://www.bungie.net/Stats/Reach/GameStats.aspx?gameid=807695504&player=PoptartGraveyrd
** Please note I’m not showing off by any means, I just found a game I played in that had two ends of the Total Value spectrum **

In this game, I played for the losing team. That means, my team’s score was 44 and the opposing team’s score was 49. I also had 20 kills, 4 deaths and 0 assists/penalties.
((20 + 0) ÷ 44) + .1 = .5545
((4 + 0) ÷ 49) + .1 = .1816
.5545 ÷ .1816 = 3.05 Total Value
I had a pretty good game, but we still lost. Since Reach is more of a team game, there was probably someone who hurt the team more than I helped. Notice the one in last at -22. 3 Kills, 25 deaths, 1 assist and 1 penalty.
((3 + 1) ÷ 44) + .1 = .1909
((25 +1) ÷ 49) + .1 = .6306
.1909 ÷ .6306 = .30 Total Value

QUICK SEGUE…
After reviewing many many games to see players who dominated, didn’t dominate, secretly dominated, etc., I’ve figured out a “rating” system for these numbers. If one would graph this, it would be shaped like a bell curve. At the pinnacle of the bell would be the “average” - “not so good”.

5.00 + = How!!
2.00 - 5.00 = MVP
1.50 - 2.00 = Solid performance
1.00 - 1.50 = Average
0.75 - 1.00 = Not so good
0.50 - 0.75 = Yuck.
0.20 - 0.50 = LVP
0.00 - 0.20 = How??

Case study #2: http://www.bungie.net/Stats/Reach/GameStats.aspx?gameid=807893287&player=PoptartGraveyrd
This case study was the reason I added .1 to both parts of the equation.

The first thing you might see in this is, well, the score. There’s no way someone from the blue team could have been MVP, right? Not so fast, buddy! If you hover over the names of the top two players on the other team, you’ll notice that they both had a higher rating than I did.

If I plug and chug my numbers into this new equation, I come up with a 7.00 Total Value. Makes sense because I had at least a part of 12 / 20 team kills without accounting for any of the team deaths.

Now If I plug and chug their team, the player who went +13 had a Total value of 2.93. Clearly a very good game. The player who went +6 has a Total value of 0.87. How? This person accounted for almost half of his team’s deaths all the while leading his team in kills. Not really a bad score, but not really good either.

Conclusion…
All in all I feel that this statistic shows what an MVP to a sports team is: the Most Valuable Player. Value shouldn’t just be determined based off of kills or deaths because assists and penalties are a part of the game too.

Also, “.1” could really be any number, but I thought it showed a good exponential curve. If it were 1, all numbers seemed to fall between 1.75 and .50, and that was really hard to separate the okay-but-good games from the okay-but-bad games. If I wanted a really large separation system, I could go with “.01” or lower.

Thoughts??

2

Problems I found:

If you calculate the Total Values for all of the players from the game from Case Study #2 based on the formula you provided, you get the following numbers (I will include their kills, deaths, and assists in parentheses and name them A to H based on their order top to bottom on the carnage report)

Red Team (50)

Player A 0.74 (17 kills, 2 assists, 11 deaths)
Player B 2.93 (14 kills, 3 assists, 1 death)
Player C 0.84 (12 kills, 2 assists, 7 deaths)
Player D 1.73 (7 kills, 1 assist, 1 death)

Blue Team (20)

Player E 7.00 (10 kills, 2 assists, 0 deaths)
Player F 0.75 (4 kills, 0 assists, 15 deaths)
Player G 0.71 (4 kills, 0 assists, 16 deaths)
Player H 0.42 (2 kills, 0 assists, 19 deaths)

*All comments made from here are based on the assumption that my math is right. Feel free to check the math before you take my comments with any kind of weight.

One odd thing that sticks out to me immediately is that the person on the losing team who went -11 with no assists (Player F) has a slightly higher Total Value than the person on the winning team who went +6 with 2 assists (Player A). The reason for this is almost exclusively due to the fact that the margin of defeat was 30. If you change the team score for Blue Team from 20 to 40, making the margin of defeat only 10 for Player F’s team, Player F’s total value decreases from 0.75 to 0.5 (ignoring all other players and keeping Player F’s stats the same) and Player A’s total value increases from 0.74 to 1.28 (again ignoring all other players and keeping Player A’s stats the same). This means that if a player can effectively help the other team do better without affecting their own personal stats (i.e. bother their own teammates, camp in a corner out of the action, leaving their teammates with less help, keeping good weapons away from their teammates, etc.), they can increase their own Total Value.

To further illustrate this, let’s make an example game and take stats from 2 players (kills/assists just means kills + assists):

Red Team (50)

Player A (21 kills/assists, 4 deaths)

Blue Team (10)

Player B (8 kills/assists, 15 deaths)

Player A’s Total Value is 1.04, while Player B’s Total Value is 2.25.

This is so because while Player A had, at least assists for, 42% of their team’s kills and 40% of their team’s deaths, Player B had, at least assists for, 80% of their team’s kills and 30% of the team’s deaths. You get the Total Values for Players A and B shown above when you, while taking into account the factor of 0.1 you add to kills and deaths in the formula, divide the percentage of the team’s kills they had a hand in by the percentage of the team’s deaths they accounted for. If you took a decent sample of games, 100-200 games would probably suffice, and measured the Total Values of each player in the game and performed a deep analysis of the data, I can tell you with certainty that you would almost definitely find that as the margin of victory for one team goes up, the likelihood of the player with the highest Total Value being on that team goes down at an exponential rate.

So, Total Value only really accounts for the percentage of positive and negative impact you had on your team’s score, which, in a team based game, might be a comprehensive enough way of measuring a player’s performance or determining who deserves to win an “MVP” award in, at most, a small majority of cases, but would be greatly flawed for a good number of cases.

However, tweaking this formula and combining it with True Skill or a similar system, which takes just team performance into account, might be a good way to make individual performance count more than it has in ranking systems in the past in Halo.

Sorry, if this entry is rambly and incoherent, but I don’t feel like reading it over again.

3

Thanks for the response! Glad someone is actually thinking about it!

This isn’t meant to be argumentative, merely to help clear things up.

I understand everything your saying at it’s something I’ve been thinking about, but I don’t know if there’s a way I can account for blowout wins/losses without adding a bland number. (ex. +1 for a win). I personally don’t think it’s fair for the close games.

> This means that if a player can effectively help the other team do better without affecting their own personal stats (i.e. bother their own teammates, camp in a corner out of the action, leaving their teammates with less help, keeping good weapons away from their teammates, etc.), they can increase their own Total Value.

You are exactly right that it could make people camp just for their own good, but that just means the whole game needs more of an impact on “wins”. I don’t know if there’s something I can plug in that plays off the value of the score, but if there’s a bonus for wins it wouldn’t be fair for the person who’s playing really well on a bad team. All in all, the closer the game, the more accurate the Total Value.

After mulling this over for a little while, I’m wondering about a few adjustments to the formula. The first would be to add a “Blowout” bonus to the winning team for a win of 20 or more. The second deals with assists and penalties. If I multiply those by something, say .75 or .5 to show that they’re not the same value as a kill. The reason I haven’t just done it is because you can’t get more assists & kills than the final score (since one person can’t get more than one assist at a time).

Thoughts?

4

MVP was given to the player with highest percentage of kills on whichever team. Also, what are you trying to use your MVP for? Ranking players or just like Halo 3’s MVP?

5

MVP system was good: Whoever slayed in slayer; whoever sacrificed in objective.
Most Valued Player.

I dont value campers.

6

> MVP was given to the player with highest percentage of kills on whichever team. Also, what are you trying to use your MVP for? Ranking players or just like Halo 3’s MVP?

Nothing special, just something similar to what was in Halo 3.

> MVP system was good: Whoever slayed in slayer; whoever sacrificed in objective.
> Most Valued Player.
>
> I dont value campers.

I completely disagree - there was no tact to getting the most kills. (The objective was fine though) While it’s true they got the most kills, that doesn’t mean they were the most valuable. You could be on the team of someone who went 20-20 in a 50-49 game. What if you went 18-5? That guy died 4 times as much as you but only got two more kills? Doesn’t seem fair to me. Value is a two-sided affair.

And also, the only thing worse than a camper is a rusher.

7

You could create a multiplier that is contingent on whether or not your team won the game and by how much, but that still doesn’t seem to be enough to solve the inherent problem. I really don’t see this as a system that would work standing on its own, but combining it, after tweaking it quite a bit of course, with True Skill or another ranking system that focuses solely on team performance, might provide a good balance in a ranking system between team and individual performance. You might be on to something here that you weren’t trying for.