When Battle Points Attack, ...err Fail!

I want to share a scenario I worked up based on a standard 3 game battle points driven tournament. You'll see some interesting results from the tournament format favored by many 40k players.

A lot of other bloggers have posted conclusions similar to mine, but none have shown any examples of how the system fails.


Just to be clear this scenario was constructed specifically to show the deficiencies of the battle points system. It is possible that this can occur in a real world situation. We'll be looking at battle points alone, we'll assume that every player is a gentlemanly golden daemon winner. Also for this example we've cobbled together a magic 8-ball, an e-meter, and a Ouija board to create a device that can accurately determine a 40k player's skill level between 0 and 10. When two players play each other an equal skill level results in a draw, a difference of 1 results in a minor victory, and a difference of 2 or more results in a major victory. The games are scored at 8 points for a major, 5 for a minor, 3 for a draw, and a 1 for a loss. There will be a random first round pairing which we are going to force into a purposely bad situation. It's a 3 game 16 player tournament with players and skill levels as follows:

Player	Skill
A	3
B	0
C	6
D	0
E	4
F	1
G	2
H	0
I	5
J	4
K	7
L	8
M	9
N	9
O	9
P	10

In this scenario the best players where accidentally seeded to play each other in the first round. A few mediocre players gained initial pairings with some baby seals. Parings are generated every subsequent round according to battle points. The rounds and pairings go as follows:

Round 1		Round 2		Round 3
A B	A(8)* B(1)	A C	A(8)* C(16)**	C E	C(24)*** E(17)**
C D	C(8)* D(1)	E G	E(16)** G(9)*	L N	L(14)** N(19)-**
E F	E(8)* F(1)	I L	I(6)* L(13)**	P G	P(18)*** G(10)*
G H	G(8)* H(1)	P M	P(10)** M(4)-	J O	J(10)* O(17)**
I J	I(5)* J(1)	N B	N(11)-* B(2)	A I	A(16)** I(7)*
K L	K(1) L(5)*	D F	D(2) F(6)*	F M	F(7)* M(12)-*
M N	M(3)- N(3)-	H J	H(2) J(9)*	B D	B(6)- D(6)-
O P	O(1) P(5)*	K O	K(2) O(9)*	H K	H(3) K(10)*

The expected results according to player skill should be: P, M, N O, L, K, C, I, J, E, A, J, F, D, H, B.

Instead we get: C, N, P, E, O, A, L, M, G, J, K, I, F, B, D, H.

In a situation where we know what the results should be, we see that battle points fails to get us there. Even if the tournament went 4 rounds and the two undefeated players duked it out, the tournament would leave a tie for first place. If in that situation, C had 1 more paint or sport point he would win out over someone who beat him. Even if P had a bad game and only scored a minor victory, he would still be taking home second place. This says nothing of the jumbled mess that is the rest of the bracket. K is rated as a better player than C and ends up in the bottom half of the results.

There are going to be bad paring in any tournament format. There's nothing that stops the two best players at the tournament from getting each other in the first round. However, in a battle point system you run in to the "clubbing baby seals" effect. Bad players are like rocket fuel to the scores of people who play them. Since the system rewards degree of victory, people who play games against opponents of drastically lower skill gain huge point advantages. Playing a close game with a competent opponent penalizes you.

I tried the same setup using a 4 game opposite seed pairing and got much better results. P ended up the winner, and C was in the middle of the pack where he should have been.