The 'bouncebackability' system

[Orpheus]

Re: The 'boucebackability' system Football: 20-6; +1163 NHL: 4-2; +171 Tonight NHL play: Boston 2.38 [note: the Vegas line is 2.30 and that is exactly on the 'benchmark odds' mark ... if it rises above 2.30 again before midnight (UK time), I will cancel the play here. In contrast to my football plays, the US sports plays use closing odds in the analysis, so my tips for these sports would be typically close to midnight during the week.]

[Orpheus]

Re: The 'boucebackability' system

I don't get this filter - First you find a match that should be profitable because of all the criterias mentioned, and then you add this filter that says that it must have shown profit for all matches within this odds range, otherwise it's a no-bet? Doesn't this defeat the entire purpose of the process? :unsure

I accept that this fourth stage is just a filter and so is not central to the principle of the 'bouncebackability' system, however I don't accept that it defeats the entire process. The reason for my adoption of the fourth stage is best explained using the 'HomeWin' or 'RoadWin' elements, i.e. I'm looking to lay a team that meets my criteria. I introduced this stage during the baseball season as I was uncomfortable with the fact that decent teams were winning, meeting the criteria to be laid in their next match, but that their next match was against poor opposition, e.g. Washington Nationals. The match odds were meeting the selection criterion re: benchmark odds, but I just couldn't bring myself to trust a system that would lead me to backing such poor teams. Hence, I introduced the criterion that I would only back the opposing team if it had proven profitable to back that team at the available odds over the past 100 home (/away) games. The effect is to reduce to the number of qualifying plays, but hopefully not by a large margin. If it has proven profitable to back a team following a loss at the current odds (this follows if the current odds are within the benchmark odds from part 1 - see the Celtic example), then they must have been very unprofitable to back following a win to make the overall profitability negative. You've raised a valid point about the necessity of part (4), so I've gone over the spreadsheets to see how my system results from last weekend would have changed if I had removed part (4) from the process. These are the additional plays that would have resulted: Lyon 2.10 [won] Lazio 2.80 [drew] VVV Venlo 2.74 [drew] AZ Alkmaar 2.58 [won] lay Feyenoord 7.35 [lost] Denizlispor 2.50 [lost] lay Gaziantepspor 1.73 [won] Genclerbirligi 1.83 [lost] lay Kasimpasa 2.70 [drew] Results: 4 winning plays; 5 losing plays; loss of 52pts. Only a small sample, but it would have resulted in increased exposure and no extra profit. However, I will track back to the start of the season and see if the extra plays resulting from dropping part (4) do make a postive difference to the profits over the larger sample period. Thanks for your thoughts, Hooloovoo. This is the feedback that I'm looking for when posting my system :ok

[Orpheus]

Re: The 'boucebackability' system Football: 20-6; +1163 NHL: 4-3; +71 Tonight's football play: Aston Villa 3.10

[Orpheus]

Re: The 'boucebackability' system Football: 20-7; +1063 NHL: 4-3; +71 Tonight's NHL play: Buffalo 1.55

[blerky]

Re: The 'boucebackability' system Hi Orpheus, really good start! Very interesting approach, i was thinking something similar using the “interruptions”, have you got any expected ROI in the medium/long run? I will follow with interest, good luck.

[Orpheus]

Re: The 'boucebackability' system Thanks blerky :) In terms of ROI, I don't really think in those terms as it is misleading in a gambling (as opposed to financial investment) environment. I think that the system will be worth my time/effort if it doubles my bankroll by the end of the year ... a ROTE (return on time/effort, if you like). In terms of what is therefore an acceptable ROI, it depends on the number of plays that I would expect to make over the course of a year. I'm currently averaging 40 plays per week, so that would 2080 over the course of a year. Let's say my initial bankroll is 5000pts, so with my 100pts per bet stakes (more if backing favourites), that is about initial 2% of bankroll. What ROI on 2080 bets at 100pts (for simplicity's sake) would return 5000pts profit? That's 5000/2080 = 2.40%. If my average odds are >2.0, the required ROI would be smaller; if my average odds are

[Hooloovoo]

Re: The 'boucebackability' system

Odds ....... Profit Odds of 0 - 1.70 have a "dip" in profit. Have you examined various "odds bands"? I.e, Is it possible that only taking odds between, say, 1.80 - 2.40 would yield even more profit? (it would mean less selections of course)

[Orpheus]

Re: The 'boucebackability' system There is indeed a dip in profit for the lower odds ranges, but this is only the case for the example club, Celtic. Each team has a different profitability distribution, so, if anything, it would be better to look at more odds ranges (e.g.

[Hooloovoo]

Re: The 'boucebackability' system I had a quick look at premier league for the last 4 years back in time. (05/06 - 08/09) (Wrote a small java program to help me out with the calculations) Following this (very simple compared to yours) strategy: Looking only for home matches for a specific team. When they have the following sequence: Match 1: WIN Match 2: DRAW/LOSS Back them to win in match 3, regardless of opposition. This strategy (which is a very simplified version of your "HomeLoss" version, and without the benchmark odds parameter) shows profit for a number of teams (in Premier League). BUT: The sample size for each team is very small, maybe 5 occasions per season. So over the four seasons i've analyzed, there are only a very small number of bets per team (about 15-20) So how reliable is this, i guess you must have experienced the same, as you say you analyze on a per-team basis???

[Orpheus]

Re: The 'boucebackability' system Yes, I've experienced the same, but I wouldn't call 5 of 19 home games under consideration for the 'HomeLoss' system a small number. Don't forget that if match 3 is a win, that brings the team into the 'HomeWin' analysis ... and then there is the away game counterpart for these two elements. Even after all my restricting filters, of the 107 games played in the Premier League, 18 of them have become system plays [6 'HomeLoss'; 2 'HomeWin'; 5 'RoadLoss'; and 5 'RoadWin']. I'm happy with that ratio of bets:games.

[Orpheus]

Re: The 'boucebackability' system Football: 20-7; +1063 NHL: 5-3; +171 Tonight's NHL plays: Atlanta 1.77 Vancouver 2.10 Chicago 1.80 Anaheim 1.61

[Orpheus]

Re: The 'boucebackability' system Football: 20-7; +1063 NHL: 7-5; +126 Today's football plays: Lay Blackburn 2.05 Man City 1.39 Tottenham 1.59 lay Setubal 2.74 Motherwell 2.44 Stenhousemuir 3.10 Real Madrid 2.14

[Orpheus]

Re: The 'boucebackability' system Football: 23-11; +816 NHL: 7-5; +126 Tonight's NHL plays: New Jersey 2.25 Philadelphia 1.63 Dallas 2.15 Phoenix 2.35 San Jose 1.63

[Orpheus]

Re: The 'boucebackability' system Football: 23-11; +816 NHL: 10-7; +251 Today's football plays: lay Marseille 4.00 lay Bologna 2.98 lay Maritimo 5.60 Almeria 2.60 Sevilla 1.67 Ankaragucu 1.76 Eskisehirphor 1.65 Manisaspor 2.12

[Orpheus]

Re: The 'boucebackability' system Football: 27-15; +386 NHL: 10-7; +251 Today's football play: Nantes 1.75

[Orpheus]

Re: The 'boucebackability' system Football: 27-16; +253 NHL: 10-7; +251 Today's NHL plays: Toronto 1.73 Calgary 1.85 San Jose 1.40

[Orpheus]

Re: The 'boucebackability' system Football: 27-16; +253 NHL: 12-8; +314 Today's NHL plays: Columbus 2.00 Edmonton 3.00 * both very close to the margins, so I may cancel either play if their lines move before midnight.

[Orpheus]

Re: The 'boucebackability' system Football: 27-16; +253 NHL: 12-10; +114 Tonight's NHL plays: *Florida 2.72 *Atlanta 2.50 New Jersey 2.15 Phoenix 1.72 * very close to the 'benchmark odds' margin, so play may be cancelled if line moves before midnight.

[Orpheus]

Re: The 'boucebackability' system Football: 27-16; +253 NHL: 15-11; +412 Tonight's NHL play: Chicago 1.50

[Orpheus]

Re: The 'boucebackability' system Football: 27-16; +253 NHL: 16-11; +512 Today's football plays: lay Brentford 3.55 lay Walsall 1.94 lay Yeovil 2.56 lay Macclesfield 4.70 lay Bradford 2.63 Lincoln 2.32 lay Northampton 1.71 Rochdale 2.00 [no-play if odds rise above 2.00] Brechin 1.78 lay Cowdenbeath 2.32 Stirling 2.10 Peterhead 1.85 I use historic odds from football-data.co.uk in my model and as there can be quite a delay between those odds (published in Friday afternoon) and the game time, sometimes resulting in large odds movement that would otherwise knock out the play from the system, I will now list the benchmark odds and no-play criterion if there is a possibility that this may be reached.

[Orpheus]

Re: The 'boucebackability' system Football: 32-23; -30 NHL: 16-11; +512 Tonight's NHL plays: Philadelphia 1.65 New Jersey 1.72 Florida 1.66 Phoenix 2.06

[Orpheus]

Re: The 'boucebackability' system Football: 32-23; -30 NHL: 19-12; +664 Tonight's NHL play: San Jose 1.85

[Orpheus]

Re: The 'boucebackability' system Football: 32-23; -30 NHL: 19-13; +546 Tonight's NHL plays: Buffalo 1.53 Phoenix 2.40 Philadelphia 1.87

[Orpheus]

Re: The 'boucebackability' system Football: 32-23; -30 NHL: 21-14; +597 Tonight's football play: lay Paris SG 6.20

[Orpheus]

Re: The 'boucebackability' system Football: 33-23; +70 NHL: 21-14; +597 Tonight's NHL play: Minnesota 1.77

[Orpheus]

Re: The 'boucebackability' system Football: 32-23; -30 NHL: 22-14; +697 Today's football plays: Hull 3.37 Liverpool 1.91 lay Everton 10.50 lay Ipswich 2.08 Sheffield Utd 1.73 lay West Brom 1.72 lay Swindon 3.45 Barnet 2.11 Burton 2.33 Dagenham & Redbridge 2.14 Northampton 2.30 Notts County 1.62 Shrewsbury 2.62 lay Torquay 3.10 Rennes 1.55 Hearts 2.19 lay Athletic Bilbao 7.00 lay Besiktas 3.30

[Orpheus]

Re: The 'boucebackability' system Football: 33-23; +70 NHL: 22-14; +697 Tonight's NHL play: Toronto 2.22

[Orpheus]

Re: The 'bouncebackability' system Football: 40-34; -643 NHL: 24-14; +819 Tonight's football plays: Stoke 2.34 lay Sparta Rotterdam 4.30 Willem II 2.14 lay Manisaspor 14.50

[Orpheus]

Re: The 'bouncebackability' system Football: 42-36 (one loss had been incorrectly labelled as a win); -409 NHL: 24-14; +819 Sunday clawed back some of Saturday's football losses, but it's time for a reappraisal. First, to recap my first post in this thread:

This is a system that should have some applicability to football, but its primary focus is North American sports and especially baseball, basketball and ice hockey as these sports involves teams playing many times per week.

There are three reasons why I said that this system should only have some applicability to football :

1. Because of the third game outcome - the draw - it is not as clear-cut how well teams will react to a draw; it is black-and-white with win or loss as in North American sports. I have tied them with losses ('Home Fail to Win' system rather than simple 'Home Loss' system), but maybe this third game outcome simply dampens any 'boucebackability' effect.
2. My sample for each team in North American sports is 100 home (or away) games, which spreads itself over 1 1/2 seasons for the North American sports (except NFL). In order to get a raw sample of 100 home games for football, I need about 5 seasons of past matches. The problem here is that how well a team performed 5 years ago should have very little impact on current games. This is the tradeoff: sample size vs. sample relevance.
3. Promotion/relegation. Other than franchise changes, it is always the same teams in North American sports. Not only are there sample problems as in (2), but the sample size shrinks further when teams do not stay in the same division, e.g. Birmingham's last five seasons have seen them relegated from Premier League, promoted from Championship, relegated from Premier League, promoted from Championship, and now maybe relegated from Premier League again. That's a different type of bouncebackability! The problem is even worse in the lower leagues as teams can leave a division by either promotion or relegation so there is a greater problem of finding a decent sample size, even over 5 years.

That's why I was amazed to see how well the system had started (19-7; +1163 in the first weekend) when I had only been hoping to "find an edge". There is little that I can do about problem (1) as ignoring games that end in a draw is not a feasible option and I can't see any immediate solution to problem (2). Anyone else got any suggestions? That leaves problem (3). As such, I have broken down the plays that I have listed in this thread according to tier (Premier League = tier 1; Championship = tier 2, etc.). Here are the results: Tier 1 Won: 26; Lost: 17; Profit: +159 Tier 2 Won: 5; Lost: 2; Profit: +320 Tier 3 Won: 5; Lost: 7; Profit: -469 Tier 4 Won: 6; Lost: 10; Profit: -419 Other than some very good results (although from only a small sample) of second tier leagues, it is clear where the system has failed over this short period (the failings are not so clear-cut if the sample period begins at the start of the season, but lower league games are clearly not as well predicted by this system as higher leagues games). I will therefore only consider games in tier 1 leagues (i.e. the top division in each league) from now on as a means of offsetting problem (3). The overall profits are not very high, but at least they are positive and the large early successes of this system were almost exclusively in tier 1 leagues. I'll continue with this for football, but if it continues to not be a money-earner, then I'll go back to my initial thoughts and just apply this system to North American sports. Updated record: Football: 26-17; +159 NHL: 24-14; +819

[Orpheus]

Re: The 'bouncebackability' system Football: 26-17; +159 NHL: 24-14; +819 Today's football play: lay Blackburn 3.95