Magic_Umbreon
Researching Tower Scientist, Retired
Does anyone here map one figure against another?
E.g. for Murkrow versus Treecko in crude approximations...
Whirlwind 4; Surprise 3; Peck 2; Miss 1 = 10 for Murkrow (e.g. 2/5 chance of Whirlwind) and Slam 3; Dodge 2; Pound 1 = 6 (e.g. 1/2 I know it's a bit more actually for slam).
Then I look at each attack on a figure and the chance it has of every outcome on the opponent's figure:
2/5 * 2/3 = Whirlwind beating Treecko = 26.67%
3/10 * 1/2 = surprise losing to Treeko = 15%
3/10 * 2/3 = miss or peck losing to Treecko = 20%
The remaining is the chance of no effect due to a dodge spin or equal white spin = 38.33%
So, I would summarise Murkrow versus Treecko as:
Treecko win: 35%
Murkrow win: 26.67%
Stalemate: 38.33%
None stalemates: 59% Treecko.
For the figures I approximated for Treecko and Murkrow, the end result is that Treecko will win 59% of the none-stalemate games.
The big flaws in the method I used:
I approximated the figures because of laziness but this wouldn't happen normally;
The strengths of the purple effect is not accounted for, nor is movement speed etc. simply how they react in a battle.
I've been thinking about using this method of matching one figure's spin probablities against another for every combination of figures. That data could then be used for "teching" against popular figures, highlighting rows in a spreadsheet to show which figures will beat your entire team (to see your team's weaknesses) and by a series of methods of excluding the weakest figure each time (reflecting a "metagame" in that the figures that beat the better figures are more useful than those that beat the weaker and not used ones) could even show the potentially BEST 6 FIGURES for a team.
It's just in ideas stage now, but I plan to write some computer programs that will do what I did above when given the figure data.
It could look something like this for a subset of 3 figures:
First the data for A, B and C. The contents of the spin wheel doesn't actually matter, because what matters is the result of that data in coompetition to each other. I'd run the program that would give me data for their matchups like so:
v|A |B |C |OVERALL
A|0.5|0.7|0.2 | 0.467
B|0.3|0.5|0.9 | 0.567
C|0.8|0.1|0.5 | 0.467
This is for A, B and C being nontransitive figures: A > B > C > A. This relationship is explained on a wikipedia page.
The "overall" column clearly reflects what is quite obvious between the three figures: because B beats C and C beats A a lot more than A beats B, B is the best figure but in a way only because C is "helping along".
While the table doesn't show anything unexpected, my idea is that by extending the table to a lot more figures I could then go on to simulate different POPULARITIES and to see what effect that has on the figures.
Does anyone think this sort of maths research might lead anywhere useful? At least you've all learnt a bit about me my friends know very well I can't just spin the figures and have fun it has to be all equations!
E.g. for Murkrow versus Treecko in crude approximations...
Whirlwind 4; Surprise 3; Peck 2; Miss 1 = 10 for Murkrow (e.g. 2/5 chance of Whirlwind) and Slam 3; Dodge 2; Pound 1 = 6 (e.g. 1/2 I know it's a bit more actually for slam).
Then I look at each attack on a figure and the chance it has of every outcome on the opponent's figure:
2/5 * 2/3 = Whirlwind beating Treecko = 26.67%
3/10 * 1/2 = surprise losing to Treeko = 15%
3/10 * 2/3 = miss or peck losing to Treecko = 20%
The remaining is the chance of no effect due to a dodge spin or equal white spin = 38.33%
So, I would summarise Murkrow versus Treecko as:
Treecko win: 35%
Murkrow win: 26.67%
Stalemate: 38.33%
None stalemates: 59% Treecko.
For the figures I approximated for Treecko and Murkrow, the end result is that Treecko will win 59% of the none-stalemate games.
The big flaws in the method I used:
I approximated the figures because of laziness but this wouldn't happen normally;
The strengths of the purple effect is not accounted for, nor is movement speed etc. simply how they react in a battle.
I've been thinking about using this method of matching one figure's spin probablities against another for every combination of figures. That data could then be used for "teching" against popular figures, highlighting rows in a spreadsheet to show which figures will beat your entire team (to see your team's weaknesses) and by a series of methods of excluding the weakest figure each time (reflecting a "metagame" in that the figures that beat the better figures are more useful than those that beat the weaker and not used ones) could even show the potentially BEST 6 FIGURES for a team.
It's just in ideas stage now, but I plan to write some computer programs that will do what I did above when given the figure data.
It could look something like this for a subset of 3 figures:
First the data for A, B and C. The contents of the spin wheel doesn't actually matter, because what matters is the result of that data in coompetition to each other. I'd run the program that would give me data for their matchups like so:
v|A |B |C |OVERALL
A|0.5|0.7|0.2 | 0.467
B|0.3|0.5|0.9 | 0.567
C|0.8|0.1|0.5 | 0.467
This is for A, B and C being nontransitive figures: A > B > C > A. This relationship is explained on a wikipedia page.
The "overall" column clearly reflects what is quite obvious between the three figures: because B beats C and C beats A a lot more than A beats B, B is the best figure but in a way only because C is "helping along".
While the table doesn't show anything unexpected, my idea is that by extending the table to a lot more figures I could then go on to simulate different POPULARITIES and to see what effect that has on the figures.
Does anyone think this sort of maths research might lead anywhere useful? At least you've all learnt a bit about me my friends know very well I can't just spin the figures and have fun it has to be all equations!