Your 84 to win prediction was always unlikely Rob. We were headed for 90 points at the time. Form at the top of the table has actually dropped slightly since you made it and 88 is perhaps now more likely.
I didn't say 84 to win I said 84 for autos. You keep making that mistake and did so at the time, which was odd. At the time the league was two points behind the average, which would have been 86 for autos. The reason my prediction went wrong is because it factored in more draws. What has been extraordinary is it has been generally win or loss as between the top teams - I would have expected far more draws with top of the table clashes.
This is a bit disingenuous. I assume you are talking about the "Remaining Fixtures" thread where I asked you why you thought, I quote, "84 points may well be enough to win the league". I was intrigued because you were pretty dismissive of some other people's higher predictions even though they seemed sensible enough. I also mentioned the autos to which you replied... "That the top six are still so close this late in the season suggests that there will be a record low winning level (and, by extension, as you say a record low autos position,but we weren't discussing that on this thread)." Yup, 84 points to win, record low, not discussing autos. Got it. So how has thinking you meant 84 points to win become an odd mistake that I keep making? Irritation aside, the original point I made in that thread is that linear accrual is probably the best assumption we can make. You said it doesn't work they way but it is what we are seeing. Nothing extraordinary has happened. The top teams have simply continued accruing 1.9 points per game.
Oh come on who is being disingenuous now? Firstly you're conflating two of my predictions several weeks apart - the quote was my winning requirement not autos as I specifically said. I appreciate it confused you at the time, but surely you understand that 84 points needed to win means second place is on 83 points/84 points with a worse goal difference. Doesn't matter where first place finishes. They can have 100 points and if second place is on 83 it still means 84 points was needed to win. My subsequent prediction of 83/84 required for autos is exactly the same prediction - it means 82/83 points for third place. It was a slight a revision upwards (or at least it would have been if you had asked me at the time of the prediction to also say how much was needed to win the league), but I'm not a politician - I'm allowed to change my mind without having an internet warrior getting pernickety over time-relative contradictions. I did actually explain this on the thread - I do appreciate it's a slightly odd angle to look at it and merits explanation, but in my view as I have said on a number of occasions it's not important what total you finish on, it's only important what the guy one place behind you finishes on. That's why my predictions were phrased in an ostensibly confusing, but actually unambiguous, way. Linear accrual remains incorrect, because accrual is not linear. We've been through this. Otherwise your estimate after match day one will be that the final winner will accrue 138 points. Which is clearly ludicrous. I appreciate you'll say that's not a reasonable time to pick it, but if you take (to pick an extreme example), say, the table after 18 games, the points accrual as 1.94ppg giving an estimated total of 90 points but then after 19 games, the accrual was just over 1.84 ppg, or a total of 85 points for the table. So that's a five point fluctuation over one round! And that is the case for recently too. Just two matches ago the top of the table was on 75 points after 40 games. That gives a total of 86 on your linear prediction! And third place was going to end up on 84 points - just a point or so above my prediction for points needed for autos! To state the obvious, the linear prediction could even be inaccurate to three points by the final day of the season (e.g. if the top team loses, but stays in top place). So surprise, surprise you're just going to pick the match round that suits your point of view with the benefit of hindsight. Of course, if it ends up being halfway in between you can pat yourself on the back and say your prediction was sound, but the fact is some teams accelerate away at the start and slow down, others are strong finishers. Nobody knows. So you have to make a nuanced prediction based on matches to be played. Mine was that there would be a series of draws between the top teams which stifled points, but the accrual since my prediction has been incredible (channeled partly by our sensational form). I quite happy to admit I got the prediction wrong. But please don't patronise me or tell me I'm dishonest just because you didn't understand and got excited when you thought you could point out an error.
And breathe... Sorry chaps, I know it annoys, but I don't know how to cut that down without being called up on saying something ambiguous.
I said a few weeks ago that I felt the team that finishes 2nd would need around 87/88 points and after posting that, felt that that total may be to high. Now, it's beginning to look as though it could be above 90 points to finish 2nd, Is that a record????
Depends what you mean - highest third points total is 87, so we could beat that (but that requires neither Watford nor Norwich to lose a game and to win at least three) to have the highest points needed for autos, but the highest second-placed total is 92 (i.e. without winning the league) I believe. The average needed to be champion is 88 points (i.e. second-place averages 87) so were about normal for that, but the average for autos is 83 (i.e. third place averages 82) and we're way ahead of that.
Thanks for that Rob and by the time we get to 5pm on Saturday,we may have a better idea of how it's all going to finish. Mind you, I still think at least one of the top 2 positions won't be decided until the final day.
I suggested linear accrual when just 15 games remained and I don't expect it to be accurate even then, because that's an impossible goal. The question is whether it is accurate relative to other approaches. The rest just obfuscates a simple exchange. You told me I was mistaken in saying you predicted 84 to win; I got irritated and showed that you did. That was all
Well as I said if by you saying I said 84 to win, you mean that I meant 84 would be the score to beat then you'd be correct. But you keep on deliberately skewing it to be 84 points as the winners total, which it quite clearly wasn't. As I said. And I explained why. And it's also clear where my predictions went wrong - I expected draws. Otherwise I'm not sure what point you're trying to make other than "nernerner you were wrong". But then that's the point of a prediction - you might be wrong. However you clearly haven't grasped the point of the linear prediction fallacy - as demonstrated by your arbitrary 15 games remaining comment. What makes 15 games remaining better than six games remaining which was significantly closer to my prediction? I called it "So surprise, surprise you're just going to pick the match round that suits your point of view with the benefit of hindsight". See I can do nernerner too
accrual is essetinally a linear relationship in football, the slope of the regression line will be close to linear and the intercept always begins at zero
You've misunderstood my comment. There were 15 games left when I posted to the thread suggesting that linear accrual would take us to 90 points. It has nothing to do with picking a match round with the benefit of hindsight. There's no "surprise, surprise". For sure the prediction could have been revised week after week. You always want to use the most data available and if there had been 14 remaining I would have used that. I've no interest in skewing the results. If the prediction proves way out it would be equally interesting to me. You do 'nernerner' very badly.
I entirely understand your point. It's just wrong. You entirely misunderstand my point. Fifteen games to go is entirely arbitrary. What makes fifteen games to go a better linear prediction than six games to go To take your example, why not run it again after fourteen games to go and get an entirely different linear prediction? That seem is silly. That means you're just arbitrarily holding on to fifteen games to go.
