Warrior General Games

(DISCLAIMER: I did not invent Beatpaths and I do not own the idea or any of the algorithms used here--I've simply did some reverse-engineering of the code. All credit goes to the good people at www.beatpaths.com.)

Text file glory: http://www.mornacale.com/beatpaths/OProPR9.txt

Oceania Pro
Week 9
-------------------------------------------------------------------
Rank | Team Name | Record | BeatPower|Change
1. West Scranton Invaders (9-0) 100.0 +0
2. So Cal BDC (8-1) 90.0 +2
3. Banzai Pipeline 'Nites (8-1) 88.0 +5
4. Mean Machine (8-1) 80.0 -3
5. Tonga Ikale Tahi (8-1) 75.0 +2
6. Cascade Cajun Crushers (8-1) 68.0 -1
7. Orlando Prime Timers (7-2) 66.0 +3
8. Brisbane Heat (8-1) 89.0 -7
9. Mata-Utu Voodoo (7-2) 76.0 +2
10. Playboy's Players Club (6-3) 68.0 +2
11. Vanuatu Kava Sharks (6-3) 66.0 -2
12. Cook Island Convicts (7-2) 55.0 -7
13. Fakaofo Bomb Squad (4-5) 50.0 +1
14. Samoa Gangsters (5-4) 13.0 +3
15. Melbourne Thorny Devils (5-4) 38.0 +1
16. Papua New Guinea Cannibals (6-3) 23.0 -3
17. Firmament Mystic Prophets (4-5) 20.0 -2
18. Brisbane Gods (2-7) -10.0 +1
19. Barossa Grapestompers (2-7) -25.0 +3
20. Canberra Strike (4-5) -5.0 -2
21. Sth Queensland Crushers (4-5) -20.0 -1
22. Fiji Sea Turtles (3-6) -47.0 -1
23. Fiji Stingers (3-6) -62.0 +0
24. Melbourne Dreadnoks (1-8) -68.0 +1
25. Brantford Mustangs (1-8) -91.0 +1
26. Tasmania Swarm (2-7) -76.0 +1
27. Jersey Wildcats (1-8) -80.0 +2
28. Port Moresby Demons (3-6) -84.0 -4
29. Brisbane Bandits (3-6) -92.0 -1
30. Christchurch Knightmares (1-8) -96.0 +0
31t. Aussie Cardinals (0-9) -100.0 +0
31t. Emerald City Flying Monkeys (0-9) -100.0 +0

Introductory post (with information on Beatpaths): http://goallineblitz.com/game/forum_thread.pl?thread_id=1775585

Commentary
1) Look at all those busted-up ties!

As always, please ask questions and point out anything that looks buggy!

I have no time to do matchups this week, so I'll have to do them retroactively again.

Last Week: 12-4
This Season: 50-14

Beatpaths Predictions - Week 10
Home Team | Away Team | Confidence
Christchurch Knightmares vs Brisbane Heat | 185.0 - CORRECT
Cascade Cajun Crushers vs Brisbane Bandits | 160.0 - CORRECT
Firmament Mystic Prophets vs Emerald City Flying Monkeys | 120.0 - CORRECT
Jersey Wildcats vs Melbourne Thorny Devils | 118.0 - CORRECT
Fiji Stingers vs Cook Island Convicts | 117.0 - CORRECT
So Cal BDC vs Sth Queensland Crushers | 110.0 - CORRECT
Mean Machine vs Canberra Strike | 85.0 - CORRECT
Banzai Pipeline 'Nites vs Samoa Gangsters | 75.0 - WRONG wooooooo
Fiji Sea Turtles vs Aussie Cardinals | 53.0 - WRONG
Tonga Ikale Tahi vs Papua New Guinea Cannibals | 52.0 - WRONG
West Scranton Invaders vs Orlando Prime Timers | 34.0 - CORRECT
Brantford Mustangs vs Melbourne Dreadnoks | 23.0 - WRONG
Playboy's Players Club vs Fakaofo Bomb Squad | 18.0 - CORRECT
Barossa Grapestompers vs Brisbane Gods | 15.0 - CORRECT
Mata-Utu Voodoo vs Vanuatu Kava Sharks | 10.0 - CORRECT
Tasmania Swarm vs Port Moresby Demons | 8.0 - CORRECT

This Week: 12-4

Only one upset outside of the 55-or-under upset zone.

Originally posted by Mornacale
14. Samoa Gangsters (5-4) 13.0 +3
15. Melbourne Thorny Devils (5-4) 38.0 +1

This does not compute. 38 > 13.

There's also OPT over Brisbane. I'm curious as to how this works out.

Originally posted by captobvious
Originally posted by Mornacale

14. Samoa Gangsters (5-4) 13.0 +3
15. Melbourne Thorny Devils (5-4) 38.0 +1

This does not compute. 38 > 13.

The issue is that those teams are not actually tied, so the tiebreaker doesn't matter. I'll use Moose's example, since it's easier to work through. The issue is that OPT has (well, had) no Beatlosses. OPT => MM => FBS => TIT => OPT gets rid of one, while OPT => MM => PPC => CCC => OPT gets rid of the other. So, Orlando gets put in the first "tier" of teams (those with no Beatlosses). On the other hand, the Heat do have a Beatloss, to BDC, so they get ranked below all the no-Beatloss teams regardless of tiebreaker.

Moving down a little bit, Samoa clearly has at least one Beatloss, but all of their Beatlosses get removed from the graph before all of MTD's, thus letting Samoa get ranked higher.

Movin' on up.

Originally posted by justinkicksass
Movin' on up.

Sadly, I don't think you will be favored to win this next game.

Originally posted by Mornacale
Originally posted by justinkicksass

Movin' on up.

Sadly, I don't think you will be favored to win this next game.

yeah, well that's no insult. I have SC ranked VERY high in the overall

Originally posted by Mornacale
I have no time to do matchups this week, so I'll have to do them retroactively again.

Last Week: 12-4
This Season: 50-14

Beatpaths Predictions - Week 10
Home Team | Away Team | Confidence
Christchurch Knightmares vs Brisbane Heat | 185.0 - CORRECT
Cascade Cajun Crushers vs Brisbane Bandits | 160.0 - CORRECT
Firmament Mystic Prophets vs Emerald City Flying Monkeys | 120.0 - CORRECT
Jersey Wildcats vs Melbourne Thorny Devils | 118.0 - CORRECT
Fiji Stingers vs Cook Island Convicts | 117.0 - CORRECT
So Cal BDC vs Sth Queensland Crushers | 110.0 - CORRECT
Mean Machine vs Canberra Strike | 85.0 - CORRECT
Banzai Pipeline 'Nites vs Samoa Gangsters | 75.0 - WRONG wooooooo
Fiji Sea Turtles vs Aussie Cardinals | 53.0 - WRONG
Tonga Ikale Tahi vs Papua New Guinea Cannibals | 52.0 - WRONG
West Scranton Invaders vs Orlando Prime Timers | 34.0 - CORRECT
Brantford Mustangs vs Melbourne Dreadnoks | 23.0 - WRONG
Playboy's Players Club vs Fakaofo Bomb Squad | 18.0 - CORRECT
Barossa Grapestompers vs Brisbane Gods | 15.0 - CORRECT
Mata-Utu Voodoo vs Vanuatu Kava Sharks | 10.0 - CORRECT
Tasmania Swarm vs Port Moresby Demons | 8.0 - CORRECT

This Week: 12-4

Only one upset outside of the 55-or-under upset zone.

is the (10) the point spread?

the (10) is the confidence that beatpaths has in the pick. a higher number means the teams are separated by more on the beatpaths and are, in theory, more likely to win.

the 10 means that beatpaths pretty much said "wtf someone has to win but this should be close"

gotcha..ok,

Moose pretty much has it. To be technical, though, the confidence actually has nothing to do with predicting the score. It's more of a measure of how certain we can be that the one team is better than another.

For example: imagine that Team A is a team which has had every single one of its Beatwins and Beatlosses looped away (this actually happened to the Denver Broncos at one point this season). Now, imagine that Team B has been involved in no Beatloops, and thus has both Beatwins and Beatlosses. Furthermore, there are an equal number of teams with Beatpaths to Team B as teams that Team B has a Beatpath to. Both of these teams are probably close to .500, though all we really know is that they each have at least one win and at least one loss.

Now, because Team A has no Beatlosses, it will go near the top of the Power Rankings. For the Week 9 rankings here, Team A would bump the Heat and take over 8th place (and yes, it is a weakness of the algorithm that a team adrift gets slotted into a high spot). Team B, however, does have Beatlosses; I would expect it to wind up right in the middle, maybe the 16-18 range. Therefore, just looking at the Rankings, it would seem that Team A is heavily favored over Team B. However, both teams have a BeatPower of 0.0. Thus, though Beatpaths would predict a win for Team A, the confidence would similarly be exactly 0.

What it is saying is this: despite their different locations on the chart, the proof that A is better than B is very inconclusive. The algorithm has a feeling that Team A is better than Team B, but it's basically just a guess, and thus there is probably only a 50/50 shot that the prediction is correct. Either way, because the algorithm lacks information on how good a team is, that also means that we really have no way to predict a final score: perhaps Team A is amazing, and blows out Team B; perhaps they are evenly matched, and have a close game; or perhaps Team B is much, much better than Team A and destroys them.

Conversely, a high confidence says that the algorithm has a great deal of evidence to suggest that Team A is better than Team B, but this doesn't mean that Team A is a lot better. Maybe Team A would win 9 out of 10 games, but win every one by a field goal. (It is likely, though, that a high confidence correlates with a high margin of victory).

Finally, keep in mind that confidence is relative. A matchup of a 1-0 team vs an 0-1 team will have 200 confidence, but even a maximum level of confidence is pretty low after just one game--sure, 100% of the evidence points that way, but 100% of a small amount is still a small amount. On the other hand, I'd say that a 15-0 vs 0-15 200-confidence matchup is pretty much a slam dunk.

tl;dr "Confidence" is a measure of how much evidence the algorithm has that one team is better than the other, and is not intended to provide a prediction of game score in any way.