Let's just start by saying that nit picking is pointless because the data you're using was wrong. I've not been rude to you, or said that you've done anything deceitful, and I can understand why you used the Google data. I am sorry that it wasn't accurate enough for the purposes, and frankly I was quite surprised to see errors like the Spurs goal.
My first post was a general one about people complaining that our opponents got lots of injury time in games where up till we scored, plenty of people were praying for lots of injury time. I do believe that's true. As you replied in detail, I did then look at your data.
I first spotted that the huge difference was actually very minor. The 30% effect would disappear with one goal in one game. It's not an uncommon error - it's one I've seen before many times in reports, where something can appear statistically significant, but actually the effect is illusory, and disappears with a very minor change. I once worked with a charity which had introduced multiple new schemes because ethnic minorities were underrepresented amongst their clients by a massive 50%. I read the research, and that 50% shortfall amounted to 1 person from an ethnic minority instead of 2 being randomly selected for a survey during one month.
Up to this point I'm assuming you're using correct data. I apologise for not realising that you hadn't until I watched a little more of the Chelsea match. At which point, I realised the numbers in the actual match weren't the same as the ones you had used. If you think that I should have read your first post and assumed that the figures were wrong, then that's up to you - but I don't suspect you do.
So I thought I'd look at a handful of other matches, and discovered that the key matches which would have most effect on the stats were all recorded incorrectly. The effect of these errors were all contrary to your conclusions, so I thought you might be interested.
Honestly, I don't know why you're not interested in that, given you were interested enough to put together the first post, but if you'd rather argue the toss with me, that's fine :)
The way your first few posts come across is that you made an incorrect assumption about the number of games where we took the lead late on in the game, and when it was pointed out that you were wrong you doubled down on that and started trying to prove a case. If that wasn’t your intention, my bad. (But you must admit you were balls deep into your position before you actually looked at any data.)
Had your posts appeared the other way round I dare say I would have been more interested but as I say you came across as though you had started with your conclusions and worked backwards. I’m simply not interested in that. The reason I set out the raw data and the source material (including identifying issues I had not looked at) is so that anyone interested could check it for themselves, not so I could reject anything in the manner of a peer review out of hand. Since your first post was to say that the entire thread was a waste of time, forgive me for viewing your initial contributions as being less than wholly objective.
My main observation is that you seem to be assuming errors across the board on the part of the google data based on one instance where the explanation is probably as simple as the ball hitting the back of the net at 89:30 but the goal being awarded (or confirmed if you prefer) after the inevitable VAR check. I remember thinking at the time the goal wouldn’t stand because the Spurs player seemed at the time to come through the back of Ake so I suspect the check took at least 30 seconds. I have not checked any other sources (Sky/BBC/official PL data etc) to see if any time the goal at 89 minutes as you do. Perhaps you have, or will do.
More importantly, what is important is that you don’t mix and match your data. I have taken Google timekeeping as the starting point because their approach to any particular issue, eg whether the time of a goal is recorded as when the ball crosses the line or when it is awarded, is likely to have been adopted and applied consistently throughout the sample. If you start picking and choosing which bits of the raw data you will make adjustments to and which you won’t the analysis loses its integrity. You might as well make it up.
I would accept that in principle a small sample size means that outliers are more likely to skew the overall analysis, but an average taken across 15 games (the 16th on Sunday, incidentally, being entirely consistent with that analysis ) is not going to be affected that much by a single borderline case. Given that the methodology/raw data is replicated in relation to Liverpool’s results it is even more important to apply the same methodology consistently.
In any event, it’s evident that you haven’t actually calculated what difference the spurs game (or any other) would have made. However you know exactly how I arrived at the figures I did in the OP, so feel free to work out yourself how much difference it would have made if the Spurs score was recorded at 3-2 at 90 minutes and not 3-3. I doubt it would make a huge difference but by all means work the figures through if you disagree.
By the way, if you look back at some of your posts I think you will agree that some of them do come across as rude. I’m happy to accept that wasn’t your intention.
