Alex McFarlane wrote:The reason for acceleration at Hastings is that that is what is advertised!!!
The cutting edge information I'm sure Sean was hoping for!
Taking the list that has been developed so far:
(1) Where the event has more players than the natural limit for a Swiss. A Swiss can handle as many players as you want - this isn't the problem. The problem is having more than one person on 100% at the end. For example, you could limit the entry to a section. You could award prizes based on score, rather than placing; so having five players on 6/6 is no big deal. It is a fact that acceleration reduces the rating differences between pairs of opponents in round 1. You'd have to investigate the impact of doing this over a longer tournament to see what difference it makes.
(2) Enabling Norm seekers to meet stronger opposition than they would otherwise. Does this actually help? It may well give you stronger opponents in round 1, but using probability, I'd need to see some evidence that the norm conditions improve by acceleration. I would imagine one criterion would be increasing the average rating of your field. I suspect that the longer the tournament is, the less relevant it is; and the effect of accelerating for 3 rounds over a 9-round tournament is relatively low. The French system of 7-round acceleration for a 9-round tournament would have a more positive impact.
(3) To reduce or remove the number of mismatches between players of vastly disparate standards that arise in the early rounds of a tournament where the skill range encompasses both top GMs and near beginners. I think it is purely subjective as to whether this is a problem to be solved.
(4) Another reason for acceleration could be attracting a sponsor. The more 'big' clashes you have then the greater the number of hits on the website and the more attractive the event becomes to a sponsor. I suspect that given the pairings of an Open are only known the night before, you probably would watch the following day whatever the pairings were. I suspect the names of the players in the tournament are a far bigger attraction, rather than who they actually play.
If (1) was the reason that acceleration was initially used, then perhaps this is the natural starting point for further investigation. The objective is to avoid more than 1 player scoring 100%. There are a number of ways of doing this:
- Traditional acceleration in the UK uses quarters; so you get something like 1 v 3, 4 v 2, 5 v 7, 8 v 6 in round 1. Why quarters? (Because norm seekers tend to gather in the second quarter?) You could use some other fraction, like sixths. This reduces the rating differences even more in round 1, but the impact of doing this in subsequent rounds needs to be tested. You could do this with the dummy points system, in giving the top half a dummy point.
- You could pair as is normally done in seeded knockout tournaments: E.g. 1 v 8, 7 v 2, 3 v 6, 5 v 4. While the rating differences at the top increase, the probability of a draw would increase on the lower boards. Using traditional pairings, the probability of the higher seed winning is approximately equal all of the way through the round 1 pairings. Using these, I suspect you would have more draws, and so you would achieve the aim of knocking players out of the 100% scoregroup. I think this would be more noticeable in subsequent rounds, because chances are that the players at the very top of the tournament would have to play each other more quickly; which is one of the goals of acceleration, when traditionally implemented.
- Floats. British arbiters traditionally agree or disagree on whether the median or top should float up, depending whether or not the tournament is "long", a term defined completely arbitrarily, but conveniently such that a long tournament is longer than a normal weekender. If you float, you will by definition have to float someone up on less than 100%. So even if the rating difference of the float pairing increases, the probability of the player on 100% retaining that pairing decreases. Equivalently, it increases still further if you float the bottom down, rather than the median down. The Dutch System, used by FIDE, already has top-up bottom-down floating, so this isn't something they need to spend time on.
I think you could try any combination of the above three to help solve problem (1).