Discussion Perfect SuperCoach

Joined
21 Apr 2018
Messages
25
Likes
106
#81
Any thoughts? Has a whopping 165k left over. If I force it to have at most 50k left over I get the following instead it trades out Z.Williams for Scrimshaw and trades in Devon Smith for Balta so throws the balance out a little bit and only has 15.8k left over. Hahaha this feels like cheating.
Actually the more I stare at it the more I like it...

1552711030049.png
 

Rowsus

Statistician
Joined
19 Mar 2012
Messages
17,511
Likes
3,946
AFL Club
Melbourne
#82
It's a really interesting approach, and a very interesting read.
Thanks for your time and efforts, Steven.
My two takes .
An approach like this seems to be near guaranteed a top 10,000 finish, but is very unlikely to hit top 1,000 as it is the unpopular successful picks that seperate the chaff from the wheat. It competes with too many "same team - near as damn it" type teams, so it becomes a raffle amongst those teams.
The guess the weight of the ox comparison is just slightly off, I reckon. There was probably a greater base knowlege amongst the guessers of the ox's weight, than there is amongst the casual SC'ers. The fact that Wigg was so popular still up to a week or two ago shows the lack of depth amongst the casual player. Wigg is still in 5% of teams.
 
Joined
21 Apr 2018
Messages
25
Likes
106
#85
It's a really interesting approach, and a very interesting read.
The guess the weight of the ox comparison is just slightly off, I reckon. There was probably a greater base knowlege amongst the guessers of the ox's weight, than there is amongst the casual SC'ers. The fact that Wigg was so popular still up to a week or two ago shows the lack of depth amongst the casual player. Wigg is still in 5% of teams.
Yeah I totally agree. I talk a little bit about that in the article but you have put it really nicely here.

My overarching goal however is to plug in projections into the thing I was using to determine the optimal team and get a fully automated side. The problem with that however is clearly it is difficult to make any sensible forecasts for players who have never played a single game so you would definitely need to incorporate wisdom of crowds in there somehow and at the moment I can't get any better data than just total percentage ownership. Maybe if I just do that for rookies / mid pricers and then use forecasts for premiums.

Also @freowho. I'm interested. When you say "players that might be worth betting against" is that only referring to the premiums or the mid-price / rookies too? Say if I subbed out Cripps, Dangerfield, Heeney who both have 40-70%+, fro say Macrae and Lachie Neal and Tim Kelly, who are more 10-20. Would that partially satisfy your anti-establishmentarisnismness? Or am I just fluffing around at the edges?
 

Nathan

Best and Fairest
Joined
28 Dec 2012
Messages
3,308
Likes
824
#86
Actually the more I stare at it the more I like it...

View attachment 7348
Looks really good.

I would duel position switch Moore & Burgess, bench Burgess for Clarke and get Petuccelle on bench for Moore.

I disagree with Rowsus that the side has a ranking ceiling due to them all being high % players, being ultimate cookie cutter is POD in itself.
 

stephen

400 Games Club
Joined
20 May 2014
Messages
2,256
Likes
1,180
AFL Club
St Kilda
#87
Looks great. If you switch Burgess and Moore I think it makes better use of the rookies (Clark on field instead of Petrucelle).

Re Rowsus's point about it being likely for top 10,000 but unlikely to rank really high, I think that's avoidable if you changed tactics for your trades. If you planned to trade in the most popular choices each week, you're likely to stay in the fat part of the leading pack, but if you changed it up you might be able to go higher.

Do you have a plan for how to decide trades?

Will you include a 'must be listed in RD1 teams' proviso? I reckon you'll need it. The crowd is too slow to react. Maybe with an exception for the R3 floating donut.
 

Jim's Dim Sims

Rising Star Winner
Joined
29 Jan 2017
Messages
641
Likes
549
AFL Club
Collingwood
#88
Looks really good.

I would duel position switch Moore & Burgess, bench Burgess for Clarke and get Petuccelle on bench for Moore.

I disagree with Rowsus that the side has a ranking ceiling due to them all being high % players, being ultimate cookie cutter is POD in itself.
Interesting observation. A bit like this riddle.
 
Joined
21 Apr 2018
Messages
25
Likes
106
#90
Looks great. If you switch Burgess and Moore I think it makes better use of the rookies (Clark on field instead of Petrucelle). Do you have a plan for how to decide trades? Will you include a 'must be listed in RD1 teams' proviso? I reckon you'll need it. The crowd is too slow to react. Maybe with an exception for the R3 floating donut.
& @Nathan, thanks for the tips :).

That's a good idea about "must be listed in RD1 team" proviso - R3. I've looked for that data in the past when I scrape the site and haven't been able to find it but surely its there somewhere. Would you recommend all greens or accept the rookies who are named as emergencies?

In terms of trades during the season I want to plug projections into the optimal team solver and tell it what it thinks I should do. We did this last year and it worked relatively well - found Angus Brayshaw very early on but always wants to use the maximum amount of trades every week so still need to think more about it. Also the projections are pretty terrible so I want to tweak them a bit but yeah will use the first few weeks of the season to come up with something for this... future Steve's problem.
 

freowho

Dual Best & Fairest
Joined
27 Jan 2014
Messages
3,703
Likes
1,099
AFL Club
Fremantle
#92
Yeah I totally agree. I talk a little bit about that in the article but you have put it really nicely here.

My overarching goal however is to plug in projections into the thing I was using to determine the optimal team and get a fully automated side. The problem with that however is clearly it is difficult to make any sensible forecasts for players who have never played a single game so you would definitely need to incorporate wisdom of crowds in there somehow and at the moment I can't get any better data than just total percentage ownership. Maybe if I just do that for rookies / mid pricers and then use forecasts for premiums.

Also @freowho. I'm interested. When you say "players that might be worth betting against" is that only referring to the premiums or the mid-price / rookies too? Say if I subbed out Cripps, Dangerfield, Heeney who both have 40-70%+, fro say Macrae and Lachie Neal and Tim Kelly, who are more 10-20. Would that partially satisfy your anti-establishmentarisnismness? Or am I just fluffing around at the edges?
I'm far from a mathematician (in fact probably a bad one) and don't necessarily gamble from the perspective of data or percentages. I'm trying to forecast more than the "crowd" based on "football" information with an eye on the data. I'm also not interested in gracefully finishing in the top 1000 so am fully prepared to crash and burn.
For example I'm betting against 4 of your rookies but that has nothing to do with starting price or being named rnd1 which I think most of the crowd would be basing their selections on. I think 3 won't keep their spot long enough and 1 will just rot on the bench but that is the gamble I'm willing to take.
I'm also betting against 8 of your premiums. For example the data would suggest betting against Danger at his age, games played and years in the system is a good idea. The football information would suggest he has had a good preseason and will spend more time in the midfield and score better and I am more inclined to follow the football information and pick Danger. The data would suggest that at his age, games played and years in the system that Cripps is a safe pick. But the football information might suggest he is a player that relies on congestion and the AFL are trying to reduce congestion with the new rules, and Carlton have some younger players they would like to give more midfield time to. At $648K I'm prepared to bet against him based on the latter and look to trade him in if the price drop eventuates.
As I'm making decisions based on unknowns it probably doesn't help your computer program. But your computer program helps me shape some of my decisions.
 
Joined
25 Jul 2012
Messages
12,938
Likes
5,588
AFL Club
Collingwood
#97
The answer is the first one. It is special for not being a special. Every other one has something that separates them from the rest.
so just the first one ? and not the other 4 that could all be said to be different as well ?

good lord picking my team might be easier
 

Jim's Dim Sims

Rising Star Winner
Joined
29 Jan 2017
Messages
641
Likes
549
AFL Club
Collingwood
#98
so just the first one ? and not the other 4 that could all be said to be different as well ?

good lord picking my team might be easier
Every other one has a feature that separates so yes, they could be considered the odd one out. The reason it is the first one is because it is not the same kind of special as the other 4. Whereas the other 4 are special for what they have, the first one is special for what it doesn't have so is the odd one out. I hope this makes sense.
 

Darkie

Vice Captain
Joined
12 Apr 2014
Messages
5,376
Likes
2,053
AFL Club
Collingwood
#99
Every other one has a feature that separates so yes, they could be considered the odd one out. The reason it is the first one is because it is not the same kind of special as the other 4. Whereas the other 4 are special for what they have, the first one is special for what it doesn't have so is the odd one out. I hope this makes sense.
It does indeed - which makes this basically a perfect analogy. :geek:
 
Top