Wednesday 19 February 2014

Reply to Vidar.........

I received a comment today on the post where I replied to some of the criticisms Steve had made on a blog post. Here’s the comment in full:

I am following 21 UK-systems with a 150 point bank (should be 200, but I took a higher risk). I am staking 1 point on homes, and 0,5 point on aways and draws:

6-21, 6-22, 7-21, 7-22, 8-21, 8-22
31-41, 31-42, 32-41, 32-42, 33-41, 33-42
TOX, STOY, STOZ
D1-D6, D1-D7, D2-D6, D2-D7, D3-D6, D3-D7

I follow all the same systems as Steve do, plus 8 more. The difference in result is huge! So far my results are +10 pt. The lowest point early in the season was -100pt and the highest point was in early january with +50pt.

Have I chosen a better overall portfolio then Steve, or have I just been lucky with my choice of systems? What do you think Graeme?

Best regards,
Vidar

NB - I have amended the numbers slightly in Vidar’s comment as he originally said 19 systems and 6 more than Steve. Should have been 21 systems and 8 more than Steve based on the systems quoted and the portfolio he follows.

I’m conscious the blog and the systems have been getting a fair bit of stick from people who have made losses this season. However, as the SBC award shows, all members can’t be losing 50% of their bank like Steve this season or I suspect I’d be getting even more criticism than that coming my way already.

As many people with the service are aware, I do spend a lot of time each summer pulling together portfolio sheets which highlight historical performance and drawdowns of chosen portfolios for subscribers (based on the systems they want to follow) and therefore, I do see a lot of different ways to play the bets. I also regularly receive emails from some subscribers with updates of how their TFA portfolio is going and others show comparisons to other services they follow (doesn’t look pretty this season admittedly for TFA!) but the point is, I do have a fair idea of how many people are doing.

I don’t feel it’s right to break that trust though and I never mention on the blog or in emails to subscribers how others are doing. I try my best to not discuss how I’m doing either as not sure I have much to gain by saying whether I’m winning or losing as people expect tipsters to always win every day and if I say I’m losing, I get people saying I’m hopeless and saying I should stop and if I say I’m winning and others are losing, I get accused of rubbing it in if not every other person is winning. I learnt in the first season to keep my thoughts to myself regarding my own betting!

Anyway, the comment from Vidar is the first comment on the blog this season from someone not in a horrible situation with their P&L. Vidar makes an interesting point that I thought was worth highlighting (and no, it’s not that he is in profit!).

As Steve knows, he’s following 13 UK systems this season. Hence, Vidar only follows 8 systems more than Steve. 

As Vidar says, he’s in profit for the season and as I know, Steve’s lost over 50% of his betting bank! How has this happened?

Well, it goes back to my original post regarding Steve’s systems and he’s been very, very unlucky I think. I saw it again last night but Steve had 1 draw bet, the systems had 5 draw bets. Steve backed the only loser and made a 1pt loss last night on draws. The draw systems made a 165% ROI last night across 44 bets, Steve’s ROI was -100%. That’s a big swing!

The additional 8 systems Vidar follows have enough profits to wipe out all of the losses by Steve and get Vidar in profit. As I said in my reply to Vidar, any sort of profit is great this season and although a 10pt profit isn’t what Vidar would have been hoping for this season when he set out (6.67% ROC), given how horrible this season has been, to make any profit is good going. Basically, if you can make a profit this season when the results are terrible, I expect you can do it in any season! Break-even is great this season!

I guess the question is as Vidar asks, has he been lucky or did he have a better portfolio than Steve? 

Before I answer this, one point I want to make again is that you don’t need to follow 21, 13 or even 10 systems this season to be in profit. I think I have shown that already by the portfolios I developed for the SBC forum. 6 of the 7 portfolios with 5-7 systems are in profit and have achieved an average of 23% ROC this season. Hence, you didn’t need 13 or 21 systems to be in profit. However, what you needed was a diversified set of bets and not be too reliant on a small number of bets to do well.

By following all combined systems including the draw systems, the only bets Vidar missed out were the Euro bets. Hence, he’s picked a good sample of bets from every UK algorithm and importantly, he’s not played TFA roulette by picking which combined systems to follow. He’s taken the full belts and braces approach and followed them all! Not my idea of following TFA but then again, who am I to argue when Vidar is in profit this season following TFA and I’m not!

Hence, Vidar removed the luck aspect by following all UK combined systems. No idea how long he spends placing TFA bets (will be as long as me!) but what he has done has protected himself from a season where the returns by backing the bets that appear on the most systems have done very poorly. Imagine if the returns this season had been in line with previous seasons, he’d be raking it in whilst Steve would be clawing his way back to break-even.

Therefore, I don’t think Vidar has been any luckier than Steve this season. He’s backed the same bets exactly that Steve has backed and all he’s done is add a few more systems and boom, he’s not in the mess Steve is in.

We’ve all learnt a lot this season and I put myself into this category too (as would Steve I suspect) and as I keep saying, none of this detracts from the fact that the UK systems have been shockingly bad this season. However, losing 50%+ of a betting bank at this stage of the season is not right and therefore, it could and should have been avoided. When the systems do bad as they have done this season, you can’t expect to make a profit. I appreciate that and I’ll hold my hands up and admit the systems have been poor this season. I honestly do think 50%+ losses should not be occurring though.

The interesting thing with Vidar’s portfolio I see is the fact he stakes 0.5pts on Draws (would have been based on these being unproven this season). Hence, he’s understaked the best performing bets in his portfolio. That’s about as unlucky as anything this season and yet, he’s in profit. I keep saying it but we make our own luck at this game. It’s not lucky that Vidar is in profit and if he had more luck, he would be in more profit I suspect.


To answer the question Vidar asks, I believe he has a better portfolio overall as this season has shown and I think Steve’s been unlucky with his system selection but if you do the first thing correctly (have a good portfolio), you can remove a great deal of the luck element out of this game. I said it in my original post but Steve was lucky in the first season and unlucky this season. With the returns these systems have, not sure we want to be relying on luck each season to make a profit………….

23 comments:

  1. Thankyou for your answer Graeme. Makes a lot of sense.

    One thing I also want to mention is that following a lot of systems like I do can sound like a lot of work and a lot of bets to make, but it isn´t really. All I have to do is select my systems in the pivot table and I get the selections in no time, with stakes from 1-15 points on home and aways, and 1-6 points on draws. I believe the amount of matches to bet on might be slightly higher compared to a portfolio with few systems, but most of the "extra bets" by following such a big amount of systems is just adding a higher stake on some of the matches. And if the systems are working like they should, I believe it is these "extra bets" that will make the difference, as these should really be the best bets (appearing on several systems). This season it hasn´t worked as good as it should (missing 8-9 big winners as Graeme mentioned in an earlier post), but still the overall results is acceptable, at least in a shorter term (1 year).

    This season is my first following TFA-systems and I appreciate coming here on the blog and read what Graeme (and some comments) write about the systems. This way it is possible to better understand how it all works (at least in theory), and makes it easier for me to understand the systems and to set up a good portfolio for coming seasons.

    Vidar

    ReplyDelete
  2. Hi Vidar

    I would advise you to learn about staking methods.
    As you mention you place 1 - 15 stakes on home and aways.
    This is a very, very bad staking method! The risk is very high, so you need a big betting bank to protect you from the danger of bankruptcy.
    A better staking method would be to place the same number of stakes on every game.
    Take the home bets this season.
    There has been 152 different games advised on your portfolio. You have placed 620 stakes on these games ranging from 1 to 15 stakes. The result has been -25.41 stakes.
    If you had placed 4 stakes on each game you would have placed 608 stakes total and the result would have been +89.28 stakes, more than 110 stakes better than your actual number.
    And although you are risking the same number of stakes, your probability of losing your bank is much lower. Now you are so dependent on the ‘big’ bets and ‘luck’. Using a level staking method as I suggest every game counts the same. You will see a much smoother P/L graph.
    Kind regards
    Tage Poulsen

    ReplyDelete
  3. I am well aware of how to stake and any staking methods. I am playing a level stake to every system with 1 point a bet (half stake on aways and draws). If a bet is appearing on all 15 systems I follow (+6 draw-systems), then that mean I will have to stake 15 points on that bet, because I play all 15 systems with 1 point on each system. If a bet only appears on 1 system, or 5 systems, then I will stake 1 and 5 points.

    And where you get the 152 different games advised so far I don´t know. In total I have 688 bets so far this season, staking 1680 points.

    Very bad staking method? No, it´s not. You are wrong.


    Vidar

    ReplyDelete
  4. Don’t want to get too involved in this discussion but I can see both sides and you both have valid points.

    I agree with Tage in the sense that staking 15pts on one team and 1pt on another team isn’t likely to be an optimal strategy as it can’t really be the case that the team appearing 15 times is 15 times a better bet than the team appearing once! You would need a lot of 1pt winners to wipe out a 15pt loser and therefore, IMO, the staking plan is too wide.

    Putting a cap on the stake helps to reduce this exposure as Tage says.

    The difference in the bet number is that Tage is looking at unique bets Vidar and you are looking at system bets. Hence, if a team appears 10 times, you see it as 10 bets whereas Tage is seeing it as one bet.

    With the way results are going, any form of capping has improved the results this season for all portfolios. When the teams that appear the most times lose badly, then staking less on these has to give you more profit. Tage is pointing this out.

    As I keep saying though, everyone plays the bets their own way and although there may be ways to reduce the exposure, everyone has their own way of working at this game. People mocked me in the first few seasons when they followed one or two systems and doubled their betting bank as I sat and placed hundreds of bets and made less money. This season, those people following system 7-21 or system 8-22 must be licking their wounds as it hasn’t worked this season but it worked the last 3 seasons.

    I don’t believe there is a right or wrong way to follow the service. I’ve said this to Tage before but I judge my season on how much I win at the end of the season. Over the first 3 seasons, this has averaged well in excess of 100% bank growth. Yes, I take risks with my bank (based on what I want to win each season) and yes, it may not be the easiest way to make money but even this season, I’m marginally down with my TFA bets. My average growth over the first 4 seasons remains above 100% even if I break-even or have a small loss this season.

    Based on Vidar’s strategy this season, he’d have more than doubled his betting bank for the last 3 seasons I suspect. Hence, a poor season in 4 isn’t a disaster.

    Feel free to continue your debate, I’ll butt out!

    Graeme

    ReplyDelete
  5. Hi Vidar

    My example in the post was regarding the home bets as I wrote.
    Until today there have been 152 different HOME games in your portfolio.
    There have been 314 AWAY games and 225 DRAW games.
    This gives a total of 691 games, which is 3 games from your number of 688.
    Using your staking method you have placed 620 stakes on HOME bets, 780 stakes on AWAY bets and 274 stakes on DRAW bets. Total 1674 stakes in accordance with your number of 1680.

    I think I fully understand your staking method. I suggested an alternative staking method and I gave some arguments for why I believe this to be a better staking method.

    You are of course free to disagree with me, but I think it will be more productive if you explain why you disagree. A simple ‘You are wrong’ will not get us anywhere.

    Kind regards
    Tage Poulsen

    ReplyDelete
  6. I see now that you were talking about 152 different HOME games. That´s about right. But I can´t understand your thinking. Staking 4 stakes on each game? I can´t just ignore my systems. If a game appears on all 15 systems, then I have to play it, staking 1 point on each system (=15 points stake in total). Otherwise, what´s the point choosing the systems if I am not going to follow them? And opposite, if a game only appears on 1 system, I can´t just add them into 3 systems and bet them 4 times.

    Vidar

    ReplyDelete
  7. The main point in my arguments is the risk of bankruptcy.
    I wrote about it in a note to all TFA members on Oct. 2 2013.

    Of course risking 15 stakes compared to 4 stakes will introduce more volatility.
    Graeme has in recent posts analysed the established systems. This portfolio has 11 systems, so a max bet here is ‘only’ 11 stakes.
    The portfolio has lost 87.1 points (numbers from his post on Feb. 18, 2014) and he rightly focuses on system 8-22, the ‘max bet’ system. 8-22 has lost 14.6 points, but as this appears on all 11 systems, the loss on the portfolio is 11 * 14.6 = 160.6. So had system 8-22 been break even the portfolio would be in great shape.
    Missing a few winners means so much in this case.
    Maybe we should look at the brighter side of life: when we have MORE winners than expected.
    Take the first season 2010/11.
    System 8-22 had a profit of 24.7 points from 94 bets.
    Had the system been break even there would be a loss of 11 * 24.7 = 271.7 points from these 94 bets compared to the actual results, and the result for the portfolio would have been negative!

    The next season 2011/12 saw 8-22 make an unbelievable strike rate of 55.0% betting at average odds of 2.69. This is simply not possible without a great deal of luck.

    In my opinion it is very dangerous to ‘judge a season on how much you win at the end of the season’

    ReplyDelete
  8. Hi Tage. On the point around how to judge the season, I agree it’s not the best way to judge the season, however, we are all in this to make money and at the end of the day, how much I make dictates whether or not I had a good season. I personally don’t take great solace from the fact I think break-even isn’t a bad result this season. I might be telling others that but when I started out this season, I wasn’t aiming for break-even!

    I think the underlying results have been much more difficult than the first 3 seasons and therefore, I think lots of talk about the service is bullshi* as I don’t think the results have been as bad as some are making out. Yes, it’s not been great but it has been harder than the first 3 seasons, so I don’t expect the systems to do as well as the first 3 seasons. Some continue to ignore this and think my ratings should be able to adjust to short-term trends but they are all based on a long-term view of profitability and not what has happened this season.

    I think there are lots of ways to judge a season for a service like TFA but at the end of the day, subscribers are the judge and I’m happy with this as I keep saying!

    Graeme

    ReplyDelete
  9. Why do I think it is dangerous to judge a betting system by results?

    First, you don’t want to use a betting system that is not profitable. The vital question to be asked on any betting system is, if the results are due to skill or luck. This question is answered by performing one or more statistical tests on the betting record.

    Now if you don’t make those tests, you really don’t know if a positive profit could just be due to luck.
    Take system 7-22 as an example. I understand that many subscribers have this system in their portfolio.
    Before this season 7-22 had a profit of 38.75 on 220 bets (ROI 17.61%)
    This season the result is -16.22 (ROI -55.93%)
    I doubt that many subscribers have performed the necessary statistical tests and TFA has not published any. (It’s actually quite simple using a spreadsheet like Excel)
    Being a statistician myself I have performed those tests and I can report, that there is no significance for the good results in the first three seasons being the result of skill. This means, that you should not be surprised to see a loss on the system for this season.
    Actually there is not any combined system that passes the tests.
    This is the first danger. Steve is a prime example of this.

    Secondly, if most positive results are ‘luck’ and not skill, then the drawdowns are ‘incorrect’ and betting banks constructed using these drawdowns will be too optimistic, which in turn results in overstaking when hard times hit.

    NB For the interested reader here is a link to an article related to this post: http://www.football-data.co.uk/blog/luck_skill_sports_betting.php

    ReplyDelete
  10. If we wait until systems and tipsters showed statistical significance,we would be waiting a long time.

    What other tipsters do you follow which show statistical significance.

    Andrew

    ReplyDelete
    Replies
    1. Andrew,
      Actually I don’t follow any tipster at the moment.
      My interest is primarily in the theoretically aspects of betting, especially staking theory and model building.
      I’m working on a football model using the most recent results in the statistically world, which is moving very fast the past few years.
      The basic model is working now, but there is a long way to go before I can hope that it will be profitable.
      To be honest I don’t think I will ever succeed!
      But that does not bother me, I have great pleasure just updating my knowledge and trying it out in practice.

      Delete
  11. Hi Tage.

    Keeping this short as been busy tonight but I don’t think anyone argues that 220 bets isn’t a big enough sample size to judge a system. I’ve been telling everyone for a long time that the higher combined systems aren’t proven and don’t have enough bets to know if they have any significant edge. I don’t need to show stats, it’s common sense. We had this discussion last summer I think via the comments section on the blog.

    However, I sort of agree with what Andrew says around waiting for all systems to show some sort of statistical significance. Only a couple systems I run are statistically significant so far IMO and therefore, if everyone followed these, we’d all be backing the same bets every time which isn’t what I want.

    Some of my systems would takes years before I could decipher whether or not they were statistically significant. Not sure I’m that patient!

    You must think I’m crazy for taking a punt on the Draw systems this season and the Euros too. No live bets!

    Anyway, cheers for the comments.

    Graeme

    ReplyDelete
  12. Graeme,

    I don’t think anyone argues that 220 bets isn’t a big enough sample size to judge a system. I’ve been telling everyone for a long time that the higher combined systems aren’t proven and don’t have enough bets to know if they have any significant edge. I don’t need to show stats, it’s common sense.

    The above statement represent a serious misunderstanding. 220 bets are more than enough to decide via a statistical test if a betting system has an edge or not has an edge. So it’s obvious NOT common sense.
    Maybe my English is bad, but when I wrote:
    ‘Actually there is not any combined system that passes the tests.’
    I tried to tell, that ALL combined systems had enough bets to make a test on and NOT a single system had a statistical edge.
    In the article by Joseph Buchdahl (link in a post before) you can download a small Excel file that performs a test.
    I cite from this:
    ‘ This test is less reliable for records shorter than 100 bets.’
    This is because it uses the average odds. If you have a full betting record with all odds it’s possible to test much shorter betting records.
    So I’m sorry, both Andrew and you don’t understand that you don’t have to wait. The decision has already been made. Your systems has no statistical edge, not even the systems with more than 2000 bets and 3.5 seasons.

    Graeme, you are a gambler. So it would surprise me if you did not bet on systems without any live bets!

    ReplyDelete
  13. Hi Tage.

    I feel this conversation is going the way all our email exchanges go!

    I feel like I’m at Gamblers Anonymous….my name is Graeme, I am a successful gambler and I follow statistically insignificant systems!

    How many tipsters or systems have you come across that have met your significance tests? Not many I suspect although correct me if I’m wrong!

    No matter how many bets and the historical returns any system has, there is no guarantee of future success. I read a recent post from Joe where he discussed Bet Advisor and the fact that only one or two tipsters on there could be classed as statistically significant. How have these bets done over the last couple of seasons? Wasn’t great I think.

    You could wait a long time to find a system that you feel has no element of luck in the results at all and then quite quickly, the market could change, the edge could be gone and your system is useless.

    I keep going back to how I’ve done over the first 3 seasons but I don’t believe doubling your betting bank each season and achieving the results my systems have is down to luck! Yes, we may not have a test to prove this but we have our judgement. You put no weight on human judgement at all.

    Take the draw systems this season. Very early days but based on my experience of building systems in the UK leagues (28 profitable systems from 30 with the 2 losing systems being the Under/Over systems), surely it was more likely than not that the draw systems would do OK this season? Yes, there were no live results but my judgement was that these systems would do OK and so far, I’ve not been wrong.

    Yes, I appreciate these draw systems are nowhere near statistically proven and the results so far probably contain a great deal of luck but that’s why this is called gambling. I took a chance and I’ve been rewarded. I took a chance last season with some Euro bets and I wasn’t rewarded. That’s why it’s called gambling!

    Anyway, don’t want to get bogged down in another conversation like this again…….

    Graeme

    ReplyDelete
  14. Vidar,

    I have talked with Graeme, and he has told me, that my initial comment offended you.
    I owe you an apology. It was not intended of course.

    As English is not my first or second language I have usually problems in formulating me correct(ly?)
    So please forgive me.

    I wanted to tell you, that which staking method you use is as important as which portfolio you follow.
    You have chosen to place 0.5 stake on AWAY teams where Graemes ‘official’ reports use 1 stake.
    So even if you are betting on the exact same teams you will not have the same profit or loss at the end of the day. Because you are using two different staking methods.
    The systems and portfolio you follow only tells you WHICH teams you shall bet on. HOW MUCH to stake on these teams is determined by another factor, the staking method.

    There are of course infinitely many different staking methods, and I proposed a method which will surprise many (most) subscribers. But money talks.
    In this season your staking strategy has produced a tiny profit. My method has produced much higher returns.

    “But Tage, the reason for this must surely be the bad ratings this season.”, I can hear some people say.

    My answer is as always based on facts, numbers.
    Your staking method would have produced an impressive profit of 616 points if used since the start of TFA in 2010.
    My proposed method has a profit of 883 in the same timeframe.
    And then this method is only a simplified staking method. I have better methods.

    ReplyDelete
  15. Thanks for taking the time to comment Tage. I think your point is a good one about the profit figures. That explains why your method is better but doesn't make every other method wrong others use.

    Cheers,

    Graeme

    ReplyDelete
  16. NFL has just played their Super Bowl, the final between two teams in American Football.
    The game starts with the traditional coin toss, and many bookmakers offer odds on the outcome of this flip of a coin.
    Fair odds would be 2.00, but imagine that a bookie offered 2.05 on ‘HEAD’ as a promotion.
    Then you clearly have an advantage and you should certainly bet in this case.
    How much to bet is a staking strategy problem.
    This problem was solved in 1956 by Kelly, published in an academic paper.
    If you know the edge exactly like in this case, then you can use the ‘Kelly Formula’ to give you the ‘best’ bet amount.
    Notice that this staking strategy is the optimal even if the result of the toss is TAIL.
    If you knew the EXACT edge in a football game you could use this method or a variant of it.
    But of course it’s not that simple with horses or humans!

    My point is that, my staking method is not better than Vidars because 883 is greater than 616, but it will be difficult for anybody to use these figures as support of Vidars strategy.
    What I have stressed in these comments is, that Vidars method has high volatility and I have shown numbers that proves how sensitive the method is. Using his method on 8-22 in 2010/11 could easily turn a profit into a loss on a whole portfolio.

    For me staking starts and ends with controlling the risk.
    Today focus will be on teams like Shrewsbury, Crewe and Plymouth.
    If they all win the profit for TFA will be 146.18. If all lose or draw the loss is -62 a swing of more than 200 points on three games on a rainy day in lower leagues in UK.
    This is not exactly what I understand by controlling risk.

    ReplyDelete
  17. Interesting debate.

    First of all, sorry about my poor English.

    I agree with Tage, if you stake one point for each team that appears in any of the systems you are following, then you are overstaking, regardless how much in profit you are.

    My English does not allow me to explain in depth, but I recommend you the following book regarding staking plans and risk management: "Fixed Odds Sports Betting: Statistical Forecasting and Risk Management". The author is Joseph Buchdahl, the man behind football-data.co.uk

    Vidar asked what's the point following different systems then... Well, I am actually following systems 8, 22, 6-21, 33, 42, TOX, STOY, STOZ, E2-E7, D3, D7, D1-D6 and Value>=20%. But I don't stake 1 point for each time a team appears in one system. I stake a percentage of my bank according to the system with lowest suggested betting bank. For example, if a bet shows up in systems 8, 6-21 and TOX (suggested betting banks 50, 45 and 45) I stake 100/45/100=2,22% of my bank. My staking plan can be improved, of course, but is far better than Vidar's, Steve's and the most of the suscribers I suspect, which leads to bankruptcy sooner or later, regardless how much profitable your systems are (and I believe your systems are very profitable)

    I also agree with Tage regarding the importance of the statistical significance tests. But the tests I made last season disagreed with him, I really found statistical significance.

    Moreover, what increases my confidence in the systems beside the tests are other small details: higher systems more profitable than lower systems, combined systems more profitable than single systems, straight bets more profitable than handicap bets (and the higher the handicap the less profitable the system bet is), etc.

    Best regards,
    Diego T.

    ReplyDelete
  18. Hi Diego,
    When testing you have to choose a significance level.
    Joseph Buchdahl uses 1% and so do I.

    Before this season TFA had 1 system that was significant at 1% level and 12 systems that were significant at 5%.

    Today, even after the great run in the last week, the best system is 6-21 which is significant at the 1.82% level.

    Your staking method looks interesting and makes a lot of sense.
    Thank you for sharing with us.

    ReplyDelete
  19. "When testing you have to choose a significance level.
    Joseph Buchdahl uses 1% and so do I."

    In reality, Joseph Buchdahl won't go near a system which only shows 1%. I know this because he told me in a recent email exchange. He sets the benchmark at 0.1%.

    I don't know about you but I'll be long since departed this world before any systems I know of reach this mark! I understand Tage shares the same view if 'serious money' is to be staked although what constitutes serious money is anyones guess!

    Stewboss

    ReplyDelete
  20. Hi Tage,

    Probably your tests are better than mine, as I've taken some assumptions because of no having complete data. So I am not sure that I am doing it right, please correct me if I am wrong

    System 7-22: 294 bets, average odd 2.86, yield 20.8%

    Chi-squared test:

    I am aware that using average odds and yield to get expected and actual winning bets is not correct, but I think it gives an acceptable estimation. Also, I'll consider a 103% overround. So:

    "actual" losing bets: 169,8
    "actual" winning bets: 124,179
    expected losing bets: 194,2
    expected winning bets: 99,8

    p-value = 0,0026


    t-Test:

    I'll take s^2 = r(o-r) as an estimation of variance, being r the average return per bet (i.e. 1.208) and o the average odds (2.86). Again, I know this estimation is not accurate, the optimal would be using all the bets to get the sample variance, but I haven't got that data.

    No overround in this case

    t = (r-1)sqrt(n)/sqrt(r(o-r)) = (1.208-1)*sqrt(294)/sqrt(1.208*(2.86-1.208)) = 3.56/1.412 = 2.52
    degrees of freedom = n-1 = 293

    1 tail t distribution

    p-value = 0,00605

    Regards,
    Diego T.

    ReplyDelete
  21. Hi Stewboss

    0.1% is too demanding in this case in my opinion. The problem of testing statistical significance is the following: if 1.000.000 monkeys start tipping, 10.000 of them will achieve a 1% significance level. Those 10.000 monkeys think they are real winners and possibly they will begin a tipping service and will send their tips to sports-tipster.co.uk in order to proof them. Because of that Joseph finds so many previously "successful" tipsters that fail once start proofing. That's why I think Joseph has chosen such a low significance level

    As I said, there are other signs to take into account in order to determine the success of a tipping service, tough absolute certainty is not possible


    Diego T.

    ReplyDelete
  22. The common theme in this thread appears to be monkeys! Can we have some variation please? How about mice? They were the brainy ones in THGTTG or maybe dolphins although they left the planet because they couldn't get the odds!

    Stewboss

    ReplyDelete