Now that we’re done with the inaugural regular season of the ISL, we can look ahead to the Vegas finals this coming week. The teams are set, and they are who we thought they were: London Roar, Energy Standard, Cali Condors, and LA Current. I last tried to project the final when we only had four of the six meets completed – but now we have more information. Two more meets worth of swims to get a sense of what kinds of times people can throw down, plus a few swimmers competed for the first time in their respective derbies (notably Duncan Scott). Daiya Seto still hasn’t swum a meet in the ISL for Energy Standard, so I’m continuing to make up times for him here.
So what can I say about what will happen this coming weekend? Let’s take a look at a bunch of things and find out.
The very simplest thing to do is:
– pick the best two swimmers in each event across the season for each team
– stack the relays for each team based on their top splits
For instance, the opening event of the program (the womens’ 100 fly) would come out as:
W 100 Butterfly
1. 55.39 – McKEON Emma, LON
2. 55.65 – SJOSTROM Sarah, ENS
3. 55.78 – DAHLIA Kelsi, CAC
4. 56.39 – WATTEL Marie, LON
5. 56.41 – STEWART Kendyl, LAC
6. 56.97 – OSMAN Farida, LAC
7. 57.37 – SHKURDAI Anastasia, ENS
8. 57.68 – HINDS Natalie, CAC
And the women’s 4×100 freestyle relay would end up with London swimming an “A” relay of Cate Campbell, Emma McKeon, Holly Barratt, and Bronte Campbell and a “B” relay of Marie Wattel, Minna Atherton, Jeanette Ottesen, and Sydney Pickrem. Those relays would go 1-5 (with the “A” relay, incidentally, destroyed the world record – the current record stands at 3:26.53 and assuming 0.5 gain on relay takeoffs for simplicity, I have those four going 50.88, 50.60, 51.45, 51.46 for a combined 3:24.39… and they’re all Australian so it would count).
The nice thing about this approach to lineups is that it’s easy to do and it’s deterministic. We end up with a final score of:
1. 485.0 – London Roar
2. 407.5 – Energy Standard
3. 379.0 – Cali Condors
4. 368.5 – LA Current
Pretty solid win by London Roar, fairly close race for 3rd between the two US teams.
Let’s do some more interesting things.
First, let’s just add more times to the mix. European SC Champs just happened, so I’m going to add in those swimmers’ times into the mix for what they can do. Also I’m going to extend some knowledge about splitting – instead of just using a swimmer’s real 100 fly medley relay split for simulated split, I’m also going to consider what they went in the actual 100 fly. Likewise, throw in swimmer’s 50 times for the first round of skins. Still perfectly deterministic:
1. 482.0 – London Roar
2. 423.0 – Energy Standard
3. 373.5 – Cali Condors
4. 361.5 – LA Current
No change in places, but Energy moving up a bit. One reason? In the original approach, Kliment Kolesnikov splits 50.53 on the relay (which finishes 3rd), gets 3rd in the 50 back (23.29) and 4th in the 100 back (50.16). But add in his Euros times and now he leads the relay off in 49.0 (which wins), gets 2nd in the 50 back (22.75) and wins the 100 back (the same 49.0).
Now, not every team will just stack every relay. I’d be surprised to see the London women’s freestyle lineup I suggested above – there is no advantage to winning a relay by that much (with adding these extra times, they win by a “mere” 1.4 seconds over Energy). Rather than trying to determine every team’s relay lineup, I’m just going to pick completely randomly. What happens if we randomize all relay selection (subject to legal relays) and just run it a few thousand times?
The average scores we get are:
1. 472.0 – London Roar
2. 442.2 – Energy Standard
3. 368.3 – Cali Condors
4. 357.4 – LA Current
But more interesting is the distribution of places. Now we have some non-determinism (the relay selection), so we have to look at the distribution of places:
1. LON (99.5%), ENS (0.5%)
2. ENS (99.5%), LON (0.5%)
3. CAC (87.7%), LAC (12.3%)
4. LAC (87.7%), CAC (12.3%)
On 5,000 iterations, London won 4,975 of them. And those 25 that Energy won? They were very, very close – Energy won by an average of 3.7 points (456.5 to 452.8). Every. Point. Counts.
Let’s add some more noise. Up until now, I’m using the best time for each swimmer across the season as the time that they will definitely go in each race. That’s not going to happen. I’ll add two layers of noise: for a given swimmer, I’ll pick some percentage to apply to them for all of their swims, and then fuzz each swim by a random number. For instance, I might [randomly] say Matt Grevers is 0.4% faster this meet (he did say he was tapering), but then randomly add some smaller amount of noise to each swim. For specific details, I’m using X% of noise to select an amount of improvement per swimmer from a normal distribution with mean 0 and standard deviation X% and then fuzzing each swim from a normal distribution with mean 0 and standard deviation (X/4)%. This is somewhat, but at least not entirely, arbitrary.
With 0.2% of noise (a 0.2% improvement is a tenth of a second in a 50 second race), and keeping with randomizing relays:
1. LON (98.2%), ENS (1.8%)
2. ENS (98.2%), LON (1.8%)
3. CAC (79.2%), LAC (20.8%)
4. LAC (79.2%), CAC (20.8%)
Energy creeping up a little bit there, and Cali having not as much of a stranglehold over 3rd place. This isn’t that much noise either. What if we turn the knob up slightly… let’s say 0.5% of noise?
1. LON (85.5%), ENS (14.5%)
2. ENS (85.3%), LON (14.5%), CAC (0.1%), LAC (0.0%)
3. CAC (66.6%), LAC (33.3%), ENS (0.1%)
4. LAC (66.7%), CAC (33.3%)
0.5% of noise isn’t outlandish (Caeleb Dressel might get 3rd in the 100 fly behind Chad Le Clos and Tom Shields, but he won’t get ever 8th), and we’re up to Energy winning about one time in 6. Although the US teams are still relegated to 3rd and 4th place in almost all runs.
What about… a full 1% of noise?
1. LON (70.6%), ENS (28.7%), LAC (0.4%), CAC (0.3%)
2. ENS (63.5%), LON (27.3%), LAC (4.9%), CAC (4.3%)
3. CAC (52.4%), LAC (39.2%), ENS (6.6%), LON (1.8%)
4. LAC (55.4%), CAC (43.0%), ENS (1.3%), LON (0.3%)
London still the heavy favorite. And if London doesn’t win, it’s almost certain that Energy wins (97.2%) and London gets second (92.3%) — in those situations Energy beats London by an average of 467 – 437. But neither Cali nor LA are completely out of it. If everyone steps up, if the US teams do their relays right while Energy continues to DQ theirs… they could take the title too.
All I know for sure is that I’m looking forward to seeing what happens.