Friday, 3 October 2014

It's Not The Winning Or Losing That Matters

I mentioned yesterday how it seemed we have had new FTL leaders after every round, and I wasn't far wrong. did actually hold onto the lead for a few rounds, but we have had no less than seven different leaders to date. The others are Faifranco, Drawmaster, Mortimer, Randolph, Jamie A and Club Havana, and all are still in profit although 19.16 points has dwindled to less than one.

Another interesting observation is that the number of entries in profit has steadily increased. Just two were in the green early on, but by early September the number was up to 11 and now stands at 20 with two still to open their accounts, although one does have a long-awaited entry this weekend. Over Goalify is going for, you guessed it, Overs, in the Vitesse vs Ado Den Haag Eredivisie match tonight. We wish him luck on his debut.

Nicholas dropped me a suggestion as to why this season is behaving somewhat unusually so far, in regard to home advantage:

I notice your comments on the blog re Prem this season. Do you think this vanishing spray is affecting things ?
My first thought was that Nick was joking, but when you think about it, this may well have been an area where sub-consciously, referees let the home team get away with slightly more than the away side, an advantage that has now been nullified, but I think any such effect would be miniscule at best, and doesn't explain the results we are seeing.

With another International Break for the top leagues next week, I might actually have time to write a post about something other than the FTL. I've been meaning to share a post from a while back on the Betfair Forum, where a Shapeshifter had this to say about stats and ratings, and we are of the same opinion:
I've often said that in betting, the only people that profit from stats are the bookmakers.
Perhaps a simple view but on the surface, the past is not the best way to predict the future.
I've always taken the view that the 'context' of the stat is most important.

The light bulb for me came in horse racing when I started developing my own ratings. How quickly it becomes apparent that a horse who finished 6th can run faster than a horse that finished 2nd in another race. Even without ratings, looking at the lengths behind the winner can carry greater weight than how a horse "places", a number assigned to when it crossed the line, not how it crossed the line.
But you listen to pundits who will say a horse has placed on its last three outings, the papers and bookies also taking that stance.
Another example is tennis stats. I spent two year developing a rating that is given to a player’s performance. Again, the numbers build pictures on a player’s ability rather than their win/loss percentage.

Often, I will have a player go 3-1 in a tournament losing in the final. It becomes the name that everyone remembers. Meanwhile, another player went 1-1 against two better opponents and, in my book, gained some points in defeat and close to the player who went on. From there, it is a matter of weighing out the value against the rankings.
Tennis will hit you with a flurry of stats. But when you watch a match, again, the context then defines the weight you put on it rather than generalizing it. Break Point stats say more about a player able to defend than a player able to attack their opponent. And a solid serve percentage means less when it is over two sets. The number of players, especially ranked in the mid-Top 100 whose service percentage drops rapidly in the third set are there. Thus when against a player who pushes the points even over two sets will often falter during the crunch of a second set.

The bookmakers greatest tease for a punter is the past encounters. 3-1 will say the stat and 1.60 if you back him. But when one of those wins was in 2009 on a different surface than the match at hand, a line goes through it.
Then when you see that when they last met, the losing player was in the middle of a slump, losses flanked on both sides, again, the weight of that victory is questioned.
And in football, the 'form' stat is outdated. With International, Cup and FA rounds in between, the 'last six matches' is, again, a bookies dream. Then, obviously, who were these teams they played? A big difference between Man City and Fulham.
Scraping away at headline stats can often find value. Similar to a 'diamond in the rough'
My latest coup is looking how teams perform IN DEFEAT rather than in victory. On the weekend, Crystal Palace were expected to lose to Arsenal. But value on backing Arsenal wasn't there. The angle: Looking at the 'defeats', Crystal Palace still manage to play strongly in the first 45 minutes, their opponents needing to regroup at half-time to beat them. Thus DRAW/ARSENAL the play at 7/2.
For U.S. sports, the stats on baseball are presented in such a way that the lights will never go out in Vegas or New Jersey.
Two seasons ago, I started looking at how teams did over certain intervals of a nine inning game. It took a few twists and turns but eventually I found a strong window to rate a team’s performance and it has nothing to do with the result but a period within the game.
When I put my 'ratings' against 'the standings', the differences were there and an edge was created.

Bookies will put up 'Runs Per Game'. No one puts up 'Runs per 9 innings'. The extra innings in baseball are thrown into the stats. There are no ties in baseball. But, for me, if I am going to back a team, I want them to win in nine innings so that is the performance I look at, discounting the hits, the runs, the pitching during any extra period.
Last year, I built a picture on an NFL game (I believe it was Lions v Vikings). Every way I approached it, the Vikings came out on top. What the headlines said was that out of 11 visits to Detroit, the Vikings had only won twice. The odds and line went that way. I wagered on both and won.
A friend of mine who bets extensively on NFL created his own 'scoring' for matches. A simple system that defines points for a type of score. Examples, a field-goal merits one point. A touchdown merits 3 to 5. He also has a stat where each completed pass represents 'X' amount of yards. No minimum or maximum ("one 70 yard pass/run with two 25 yard passes can totally skew the stats"). That was when we were in university together. It is a bit more extensive now but still sticks to its roots.
I realize I've stated the obvious and, yes, stats are part of one’s research. But it is a matter of avoiding what I call 'the noise' and hype that stats can bring to an event.
Any views are, of course, welcome.
He later added these words:
My wording did come across as contradictory. But the main gist of it is that the public rely on the most basic of stats that are presented to them, usually by the bookmakers themselves. The extreme is to create algorithms and extensive filters to 'predict the likeliest outcome'. Around five years ago, with the help of a friend who (a) loves sports and (b) loves numbers even more, took the pythagorean theory as it is used in sports and took it further. It works great for baseball but no other sports, perhaps because of the number of games per season. I also created a set of basic filters that takes different weights on stats from football (soccer). I then tried it live and though the returns on 1x2 were okay, it did create a filter that triggered around eight matches a week from fifteen leagues world-wide of odds against OU 2.5. So where I may not believe in stats, I do believe in taking a different angle on all sports than research that is readily available and, as stated, placed like giant billboards in front of the betting public and repeated in a Clockwork Orange rehabilitation approach on in-play radio.
Readers of this blog will know that my opinion of League Tables for betting is that they are useless, and should be aware that for the Elo based ratings, the result is incidental. It's not unusual for a team to gain Elo points in a defeat, or lose Elo points after a win. I'm not a tennis bettor, but Shapeshifter mentions baseball, and with the play-offs underway, it might be a good time to remind you of the strategy I disclosed last year of opposing the pre-match favourite later in matches. As fizzer555 commented at the time:
Cassini may be on to something though as by the time you get to the 7th innings the main reason for the fav/dog bias may be out of the game (i.e. the SPs) but I think the market may retain some fav/dog bias (gradually reducing across the 9 innings) that may not be justified. Don't have data to prove that but over 2 or 3 innings the hitting strength of each team is going to be very similar except for a few outliers (e.g. Miami, Houston low hitting) and similarly with bullpen pitching a lot of the teams will be in a narrow performance range.
The recently ended game between the favourites (1.55) Los Angeles Angels of Anaheim and the Kansas City Royals (enjoying play-off baseball for the first time since 1985 incidentally), the Angels never led bet traded as low as 1.3 before losing in extra innings. The reason, as I mentioned last year, that the Angels were big favourites was due to the SP (starting pitcher) Jered Weaver being a bit of a star, but once he was out of the game, that advantage was gone, yet the markets consistently fail to correct as fast as they should when the SP is done. Yes, sometimes the bull-pen might still be stronger for the pre-game favourites, but when it gets to the play-offs, every team has earned its place there (over a 162 game season) and no team is shabby. The other advantage of play-offs is that there is a little in-play liquidity. On the few occasions I looked in on regular season games, there was very little action.  

No comments: