Competitions Pieman 2023 - Pre-season

[dogwatch]

Everybody's going with a pre-season thread so why not the Pieman?

It's January so my thoughts have turned to setting up for season 12. As discussed last year the format and rules seem to be in pretty good shape so there's probably not much tweakage required. However I thought I'd let you know where I'm headed and open it up for any further comments or ideas.

Here's a link to the discussion at the end of last year for those who (like me) need to refresh their memory ... read from this post to the bottom of the page for a range of comments.

I'm really only thinking about changing the bonus system for picking 5 different goalkickers. First I thought I'd have a look at what happened in 2022. Here's a chart of what I found.

The blue fill shows the number of people who picked five different players each week (out of a possible 48). These are the only ones giving themselves a chance at the bonus. On average about 50% of entrants picked a 5x1.
The black line is crucial - it's the number of different goalkickers we had each week. The yellow data point shows when a bonus was actually awarded (in only six games all year).
The orange column shows the number of bonuses awarded in those weeks.



What this demonstrates to me is that Wocka was right: 5x1s are for chumps. At least they were with the reward system that we had in place. It shows the strong correlation between bonuses and number of goalkickers (not so much of a correlation between bonuses earned and the number of people who went for a 5x1).

It's apparent that we usually need at least 9 goalkickers to have a decent chance at a bonus. That only happened in 8 games last year. However when it did happen a bonus was earned in over 60% of them. There was only one game where we had fewer than 9 goalkickers and somebody got the bonus (well done X_box_X !) And in Round 11 when we had 16 goalkickers we had a real bonanza of bonuses.

I know some will push for some sort of lesser bonus for picking only four different goalkickers but for several reasons I'm not going down that path so don't waste your keystrokes on it.

What I'm thinking of doing is making two changes:

1. The base level bonus increases from 20 points to 50. This is a guaranteed minimum and is a little more in line with the difficulty of picking five different goalkickers.
2. Any time the bonus is not won the bonus amount increases by 10 points for the following week. It will go on increasing by 10 points each week until it finally goes off, then it reverts to 50 points again the following week. When it does go off everybody who correctly picked five different goalkickers gets the whole bonus amount (i.e. the jackpot amount doesn't get divided by the number of winners). This will create a certain amount of variability as the season progresses inviting adaptive strategies. It will certainly draw in more chumps prepared to pick 5x1s as the jackpot increases. Perhaps even Wocka.

To illustrate how it would work, here is what the bonuses would have looked like if applied to 2022.

	R1	R2	R3	R4	R5	R6	R7	R8	R9	R10	R11	R12	R14	R15	R16	R17	R18	R19	R20	R21	R22	R23	EF
Bonuses given	0	3	0	0	0	0	0	0	0	0	14	0	0	2	0	0	3	1	0	1	0	0	0
Bonus Amount	50	60	50	60	70	80	90	100	110	120	130	50	60	70	50	60	70	50	50	60	50	60	70
Kelvin score	146​	200​	163​	80​	248​	120​	177​	182​	189​	156​	234​	116​	178​	207​	122​	86​	165​	238​	136​	141​	254​	333​	150​

As you can see, in quite a few games the bonus amount was a significant proportion of the eventual score recorded by the weekly winner. In 8 games it was worth more than half the Kelvin score. That should make the incentives for multiples and 5x1s a little more even.

Oh, and "five by one" is too much of a mouthful. Unless anyone can think of a better name for it, I'm thinking of calling a 5x1 selection a "Chump".

Let me know what you think. Not just about the bonus system but anything else Pieman-related.
I'll set up the official Pieman 2023 thread in early March.

 

[X_box_X]

There was only one game where we had fewer than 9 goalkickers and somebody got the bonus (well done X_box_X !)

Nice!

What do I win?

 

[Northern Dogs]

Nice!

What do I win?

A Chupa "Chump"?

 

[weltschmerz]

I'm a big fan of choosing five different players to chase the bonus, and when I did get the bonus it felt fairly minimal, so I support this change.

 

[ossie_21]

tl;dr

VDM x3
Hannan
Scott

Cheers DW

Seriously though, I don’t generally go for the 5 different goal kickers, but 50pts+ bonus would make it more tempting and mix it up a bit more

 

[Charlie Bucket]

80pts and you’ve got yourself a deal

 

[dogwatch]

80pts and you’ve got yourself a deal

No deal. You'll just have to wait for it to jackpot CB.

If it gets to 100 points you'll be chumping at the bit.

 

[theAthenian]

sorry for being negative but statistically for me as a punter 50 point base bonus is still too low
the temptation to pick a rsmith X3 at say 28 per goal is far far more rewarding (in points) compared to 50 point bonus
if your outlier kicks two or heaven forbid 3 - Caboom!
because to pick 5 winners would mean to pick at least 3 low yielding forwards.

for me the competition is still all about picking the outliers x2 or 3 each week
that said you do a fantastic job and i commend you my bulldog brother for making the season even more fun! Thankyou

 

[chez_ferret]

Hi dogwatch

This was this first year I had a crack at the Pieman and thoroughly enjoyed it. I tended to pick 5 x 1 up until the end of the season when it became apparent that, while this strategy had kept me in the top 10, I could not catch the leaders without doing something different. I like the challenge of picking 5 and agree that it should be better incentivised. I have tried to have a quick look at what changing the bonus would have done to the last 5 rounds and the final, comparing the actual highest score to the theoretical highest scores for a 3 +2 and a 5 x 1. Summary here, breakdown attached (not QA'd!).



A couple of observations to throw into the discussion:

• in these rounds there were two successful 5 x 1's. In Rd 19 that was 154 + 20 bonus (2nd highest score) and in rd 21 it was 121 + 20 bonus (highest score for the round).
• round 21 was the only round where a 5 x 1 would have been the highest possible score under the 20 point bonus system
• changing the bonus to 50 points would still leave rd 21 as the only rd where a 5 x 1 was the best possible score
• changing the bonus to + 50% would make 5 x 1 the best possible score in 4 of these 6 rounds but most of them would still be very close (3 of the 4 within 12 pts).

In one of your posts at the end of last year you mentioned a % bonus, hence the 50% and 100% loading comparisons. Personally, I like the concept of a 50% loading as it rewards a 5 x 1 that includes a roughy over a 5 x 1 of the favourites, and it seems to put the best possible theoretical scores for a 5 x 1 and a 3 + 2 roughly on par (albeit I am working from a very limited data set - I picked these rounds because the data was easiest to collate from your posts).

In any case, whatever you choose to implement I will definitely be on board for next year and really appreciate the time and effort you put in to running it.

Thanks

 

[grassman75]

I don't know DW, I say go back to the rules that were in place when I ruled the roost.

In some ways I think everyone should just pick 5 players with no multiples and have your bonus jackpot, but in other ways perhaps limit it to one player per team as a multiple to a maximum of X 3 and don't change the bonus base level of 25pts but do have it going up by 5pts every week.

I fear that there could be a blowout if someone snagged a mega bonus and you e worked so hard to even the comp up.

Just my two cents

Love the Pieman

 

[dogwatch]

Hi dogwatch

This was this first year I had a crack at the Pieman and thoroughly enjoyed it. I tended to pick 5 x 1 up until the end of the season when it became apparent that, while this strategy had kept me in the top 10, I could not catch the leaders without doing something different. I like the challenge of picking 5 and agree that it should be better incentivised. I have tried to have a quick look at what changing the bonus would have done to the last 5 rounds and the final, comparing the actual highest score to the theoretical highest scores for a 3 +2 and a 5 x 1. Summary here, breakdown attached (not QA'd!).

View attachment 1588460

A couple of observations to throw into the discussion:

• in these rounds there were two successful 5 x 1's. In Rd 19 that was 154 + 20 bonus (2nd highest score) and in rd 21 it was 121 + 20 bonus (highest score for the round).
• round 21 was the only round where a 5 x 1 would have been the highest possible score under the 20 point bonus system
• changing the bonus to 50 points would still leave rd 21 as the only rd where a 5 x 1 was the best possible score
• changing the bonus to + 50% would make 5 x 1 the best possible score in 4 of these 6 rounds but most of them would still be very close (3 of the 4 within 12 pts).

In one of your posts at the end of last year you mentioned a % bonus, hence the 50% and 100% loading comparisons. Personally, I like the concept of a 50% loading as it rewards a 5 x 1 that includes a roughy over a 5 x 1 of the favourites, and it seems to put the best possible theoretical scores for a 5 x 1 and a 3 + 2 roughly on par (albeit I am working from a very limited data set - I picked these rounds because the data was easiest to collate from your posts).

In any case, whatever you choose to implement I will definitely be on board for next year and really appreciate the time and effort you put in to running it.

Thanks


View attachment 1588461

Great stuff, cf.

Let me mull it over.

There is of course another angle to all this. Whatever changes are made have to be easy to code and to maintain (i.e. to make subsequent tweaks). So I'll be considering that aspect of it too.

 

[dogwatch]

I don't know DW, I say go back to the rules that were in place when I ruled the roost.

In some ways I think everyone should just pick 5 players with no multiples and have your bonus jackpot, but in other ways perhaps limit it to one player per team as a multiple to a maximum of X 3 and don't change the bonus base level of 25pts but do have it going up by 5pts every week.

I fear that there could be a blowout if someone snagged a mega bonus and you e worked so hard to even the comp up.

Just my two cents

Love the Pieman

Yep, also valid points. Will consider these too.

Except the bit about when you ruled the roost. That was too long ago for anyone to remember. ("Today a rooster, tomorrow a feather duster")

 

[dogwatch]

For Pieman boffins and enthusiasts...

I've had a chance to consider options for awarding bonuses and - with some inspiration from chez_ferret - I did a little analysis of how things might work out based on 2022 match results (and the actual Pieman selections made by contestants).

It has come down to considering two (or maybe three) bonus options for getting a Chump*. Whatever happens the old 20 point bonus will be chucked out.

* A "Chump" is a 5x1 selection where all 5 players get among the goals. ​

Thanks Wocka for inspiring the name Link: Competitions - The Pieman 2022​

The first bonus option is the one outlined earlier where there is a baseline 50 point bonus awarded for a chump every week but it also jackpoints by 10 points each week that no one gets a chump. So in a series of unsuccessful weeks the bonus would progress from 50 to 60, 70, 80 and so on. No limit to the jackpot amount ... until the season ends. Also the jackpot is not split. That is, everyone who gets a chump gets the full jackpot amount for that week.

The second option is one floated last year and which chez ferret has reminded me of. That is to give a percentage bonus instead of a flat points amount. The suggested percentage is 50% on top of the base score. So for example if you got a chump and your five players had 18, 23, 29, 56 and 63 points that would add up to 189 points. With a 50% bonus (rounded down) you would end up with 283 points (189+94). One advantage to this is it rewards people for selecting players with higher handicap values. Risk/reward.

grassman75 had suggested doing away with selecting multiples altogether (ie any 3x or 2x) but I'm not keen to do that. I think the range of available strategies adds interest. There are also reasons why you might go for a chump one week but a multiple combo the next week. However I do wish to narrow the advantage that going for multiples has over aspiring chumps. As it stands, Wocka was indeed right, as the following chart demonstrates.

Bear with me though. It's a busy chart:

The plots represented by lines are the "best possible" (ie hypothetical) results based on our goal scorers in each game last year. (Nobody actually registered these scores. They just illustrate what the perfect selection would have scored.)

• The Red line is the best possible score from selecting a 3x and a 2x (aka 3+2).
• The solid Black line is the best possible score for a chump using the 50% bonus method.
• The Green line is the best possible score for a chump using the 50points + jackpot method (jackpot amounts are based on how often chumps were recorded in 2022, as shown by the blue columns)
• For comparison purposes, the grey broken line shows the best possible score each week using the obsolescent 20 point bonus method.

The shading and columns show what actually happened last year in terms of weekly Pieman results.

• The Yellow shading shows the actual best score each week in 2022 (the Kelvin)
• The Blue columns show the actual highest chump score recorded each week. Note that while there were 24 chumps recorded all year, they only occurred in 6 games. 14 of them were in Round 11. In two of those rounds (R18 and R21) a chump won the Kelvin. On one other occasion it got within 10 points (R15). On the other three occasions it wasn't even close. Not a great advertisement, and the reason why Wocka was right.

My preference is to go with the 50% bonus approach. Reasons are as follows:

• It is much more competitive with the BPM (Best Possible with Multiples) scoring, often higher but sometimes lower.
• It doesn't blow the BPM method out of the water. That is to say, either approach could be quite productive.
• It doesn't have a massive peak in any game that could kill off competition for the season. Well, no more than the current BPM approach would. (The two "best possible" peaks on the graph are when Keath kicked 2 goals in Rd 11 and when Roarke kicked 3 goals in Rd 23. Nobody picked Keath but 2 people picked Roarke as a 3x in Rd 23 ... they split the Kelvin.)
• It provides added reward for the risk in taking players with higher handicaps.
• On average this system would have delivered a 63.3 bonus for the 24 people who registered a chump in 2022. That's better than the base level 50 point alternative proposed (but maybe not as high as the jackpot add-on would take it).
• Looking at the green line plot, the 50 point bonus method doesn't quite cut it in terms of competitiveness, even with a jackpot. It only started to look competitive in the long drought between Chumps (Rd 2 to Rd 11) when the jackpot approached its peak of 130 points.
• Both methods are easy enough to code and maintain. If anything, the 50% method is the easier one.

Now the astute among you will perhaps have noticed a few flaws in these arguments. For instance:

(1) most of the analysis above is based on the hypothetical best possible scores, not the reality that nobody ever gets a best possible score. So does the argument still hold when applied to "real life" results?​

(2) The impressive benefits of ANY chump bonus system only apply if all five players actually kick a goal. That only happened for 24 out of 556 5x1 selections across the 2022 season. That's only a 4.3% success rate.​

(3) The analysis only looks at potential Kelvin-winning scores (ie the weekly highs). That's nice, but the main aim is to win the Pieman (best over the whole season). Does one approach (BPM or either of the BPC options) lend itself to higher end of season scores? For instance one approach might deliver more consistent mid-level scores, even if it failed to get Kelvin-winning scores.​

(4) Behaviour will change according to the incentive system, so using 2022 data may not be valid as the proposed bonus systems weren't in place then.​

To these objections I would suggest the following answers:

(1) It's impossible to tell without replaying the whole season under the new rules. That's a bit too much of an indulgence I'm afraid. Besides, I like a bit of uncertainty. ​

(2) Yes it's true that only one in 23 attempts actually delivers a chump bonus but a similar argument could be made against those taking a 3+2 combo. Quite a few of those 3+2s delivered no score at all in 2023. If I get a chance I'll see if I can crunch those numbers too.​

(3) Now this is a tough question, a little more of that glorious uncertainty. However one thing you can be fairly sure of is that very few attempted chumps failed to record any score at all. Although they failed to get a bonus many of them would have 2,3 or 4 goal scorers which would keep the season tally ticking over nicely. Would the hit-or-miss 3+2 approach (and perhaps the other permissible variants of multiples) have delivered the same level of consistent scoring? I don't know.​

(4) True, but what more can a Piemeister do? You might just have to "suck it and see in '23".​

So while I'm leaning towards the straight 50% bonus method, here are two other options I might consider.

(i) Ban the 3+2 combo. That is, you can only have 3+1+1 or 2+2+1 or 2+1+1+1 or 5x1. This means a minimum of three different players. My gut feeling is against it, but I'm prepared to listen to arguments in favour of such a change. Perhaps if a tweak is still needed in 2024 this could be it?
(ii) Use the 50% bonus but combine it with the jackpot idea. That is, you get your 50% but a flat points value is also added depending on how long it is since the jackpot has gone off. It would start at 0 points but 10 points would be added for each week it doesn't go off. This would further strengthen the appeal of going for a chump as against the 3+2 method. It might also encourage adaptive strategies from one week to the next. Here's how it would have looked compared to the other two options (the straight 50% and the flat 50 point + jackpot) and compared to the BPM (3+2) method:



The same legend as the earlier graph applies: Green is for the best possible chump with a flat 50 points + jackpot, Black is for the best possible chump with a 50% bonus + jackpot, Red is for the best possible multiple (3+2). All of it is based on the actual goalscoring in 2022.

 

[dogwatch]

A little more analysis:

In the 2022 season 3+2 combos were selected 121 times (with over half of them coming from just four contestants). Their average score was 68.8 pts. They delivered 8 Kelvins (weekly winners) and the season's Pieman winner, but on the other hand over a quarter of the time (28%) they resulted in a score of zero. That's quite a spread.

The 5x1s were a lot more even. There were 556 5x1 combos. They averaged almost the same - 69.3 pts. They delivered only 3 Kelvins but they were the main combo used by the 2nd and 3rd placegetters in the Pieman. Only 4 out of the 556 (0.7%) returned a zero score. Much more reliable for keeping your season points ticking over.

If instead of a bonus of a flat 20 points for a Chump (5 different goalscorers) the proposed 50% bonus system had been in place, the 5x1s would have averaged a little more across the season (71.2%) and delivered 1 extra Kelvin.

Also here's what the top 5 finishers in the Pieman did. You can see that some of them relied heavily on 3+2 combos, while others went mostly for 5x1s or 2+1+1+1s:



Not everybody who relied heavily on 3+2s did well. ScragCity (who had a 3+2 combo 20 times) finished 43rd out of 48 and Charlie Bucket (who had nothing but 3+2s all season) finished 45th. This suggests that success is more to do with how good you are at selecting goalkickers, not which combo you rely on most. As it should be! The secret might be knowing which week to switch combos.

So you can do well with 5x1s or with 3+2s, or anything in between. You have slightly more chance of a Kelvin if you go for 3+2s but the 5x1s give you a steadier rate of scoring throughout the year. With a higher bonus for getting a Chump the 5x1s might become a bit more appealing.

Conclusion: In 2023 we'll change from a flat 20 point bonus to a percentage bonus for a "Chump". However in light of the above analysis I'm scaling back the bonus percentage from 50% to 40%. I reckon that should result in a pretty good balance. And there won't be any jackpot component.

 

[TiAn_]

A little more analysis:

In the 2022 season 3+2 combos were selected 121 times (with over half of them coming from just four contestants). Their average score was 68.8 pts. They delivered 8 Kelvins (weekly winners) and the season's Pieman winner, but on the other hand over a quarter of the time (28%) they resulted in a score of zero. That's quite a spread.

The 5x1s were a lot more even. There were 556 5x1 combos. They averaged almost the same - 69.3 pts. They delivered only 3 Kelvins but they were the main combo used by the 2nd and 3rd placegetters in the Pieman. Only 4 out of the 556 (0.7%) returned a zero score. Much more reliable for keeping your season points ticking over.

If instead of a bonus of a flat 20 points for a Chump (5 different goalscorers) the proposed 50% bonus system had been in place, the 5x1s would have averaged a little more across the season (71.2%) and delivered 1 extra Kelvin.

Also here's what the top 5 finishers in the Pieman did. You can see that some of them relied heavily on 3+2 combos, while others went mostly for 5x1s or 2+1+1+1s:

View attachment 1600193

Not everybody who relied heavily on 3+2s did well. ScragCity (who had a 3+2 combo 20 times) finished 43rd out of 48 and Charlie Bucket (who had nothing but 3+2s all season) finished 45th. This suggests that success is more to do with how good you are at selecting goalkickers, not which combo you rely on most. As it should be! The secret might be knowing which week to switch combos.

So you can do well with 5x1s or with 3+2s, or anything in between. You have slightly more chance of a Kelvin if you go for 3+2s but the 5x1s give you a steadier rate of scoring throughout the year. With a higher bonus for getting a Chump the 5x1s might become a bit more appealing.

Conclusion: In 2023 we'll change from a flat 20 point bonus to a percentage bonus for a "Chump". However in light of the above analysis I'm scaling back the bonus percentage from 50% to 40%. I reckon that should result in a pretty good balance. And there won't be any jackpot component.

I like your analysing my Pieman victory content but not your changing the rules I won the Pieman with content.

 

[dogwatch]

I like your analysing my Pieman victory content but not your changing the rules I won the Pieman with content.

If you pick 'em like you picked 'em last year you'll have no problem!

 

[TiAn_]

If you pick 'em like you picked 'em last year you'll have no problem!

Roarke Smith 4 eva

 

[grassman75]

I like your analysing my Pieman victory content but not your changing the rules I won the Pieman with content.

Cry me a river Tian...you were still in nappies when DW changed the rules after my dominant period....just think of it as taking third man up away or bringing in the stand rule

 

[grassman75]

Hey DW - Here's an idea for Pieman

In light of the Lions poking the bear by posting Dunkley kicking 3 goals against his current teammates (that's funny in itself) maybe we could select one ex bulldog this season and do something with their scores


We've got Dunks, Hunter, Schache, Ferg Greene that I can think of off the top of my head....

 

[dogwatch]

Hey DW - Here's an idea for Pieman

In light of the Lions poking the bear by posting Dunkley kicking 3 goals against his current teammates (that's funny in itself) maybe we could select one ex bulldog this season and do something with their scores


We've got Dunks, Hunter, Schache, Ferg Greene that I can think of off the top of my head....

Yeah, naah. Points for lateral thinking though.

Good thing we're not selling off our Bernie Quinlans these days. You could clean up on a deal like that.

 

[Floridog]

I am excited to get going on a new season. The changes all look good to me and based on sound research. Just one thing left to add, the 50% northern hemisphere bonus. Thanks for all the hard work to make this happen Dogwatch,

 

[Kermit]

I’m a big chaser of the bonus so any major upgrade on that works for me.

 

[Charlie Bucket]

Naughts x 3, Hannan x 2