2 hours ago, Truthiness said:

Ooof. That might be possible, but I'm not sure I'm skilled enough to pull it off.

One thing that I have done in the past is dump the file to excel and do specific analysis offline. Since I've just leveled up at work, I probably won't have time to do much of that this season, unless I get to take my work computer home.

(Phrase I didn't expect to hear at work:. We just confiscated 15 horses from an illegal rodeo. Now I need to figure out the budget implications.)

4 hours ago, Truthiness said:

Ooof. That might be possible, but I'm not sure I'm skilled enough to pull it off.

Well - let's not overshoot it, then. Maybe just having a pure count, as the sheet already has for Intel/Rogue/etc, may be sufficient to at least suggest if the question is worth digging further into.

What exactly do you want done? @xanderf

edit: I've mocked up a version which separates points spent on generic/unique squads.

Did you want specifics like how many points exactly spent on scatter aces? It's doable but a pita.

6 hours ago, Truthiness said:

It's easy to compare how many fleets included one or more of a ship. The formula for the ship sizes is done by checking against a list of ships, so making a formula that just counts one specific ship type is easy. How would this be compared? Are you looking for the raw number of each ship taken, or the percentage of fleets that take at least one of a ship type?

I'm basically wondering if people still use Assault Frigates, how popular post nerf Yavaris is, are Raiders taken that often, etc. Do Interdictors get love? Just wondering amounts of each ship used or if any look in trouble (Pelta, Vic, AF in North America are my guess).

Also, post flotilla nerf I'd wonder about fleets just running zero, one, or two. I'm not sure we need the full deep dive into them all

6 hours ago, Truthiness said:

It's easy to compare how many fleets included one or more of a ship. The formula for the ship sizes is done by checking against a list of ships, so making a formula that just counts one specific ship type is easy. How would this be compared?

What if I want to answer a question like "what percent of Sloane lists contained at least three non-flotilla ships", or "how many activations did the top 25% of MSU have, versus the bottom 75% of MSU"? Is that just formula-able?

I'd also like to see command fixing tracked alongside command manipulation (eg SFO, Support Officer).

I took the win in Orlando with my list:

Screed Interdictor & Rogues
Author: DaveG

Faction: Galactic Empire
Points: 397/400

Commander: Admiral Screed

Assault Objective: Most Wanted
Defense Objective: Fighter Ambush
Navigation Objective: Superior Positions

Gladiator II-Class Star Destroyer (62 points)
- Demolisher ( 10 points)
- Agent Kallus ( 3 points)
- Ordnance Experts ( 4 points)
- Engine Techs ( 8 points)
- Flechette Torpedoes ( 3 points)
= 90 total ship cost

Gozanti-class Cruisers (23 points)
- Hondo Ohnaka ( 2 points)
- Comms Net ( 2 points)
= 27 total ship cost

[ flagship ] Interdictor-class Suppression Refit (90 points)
- Admiral Screed ( 26 points)
- Interdictor ( 3 points)
- Captain Brunson ( 5 points)
- Engine Techs ( 8 points)
- Disposable Capacitors ( 3 points)
- Heavy Ion Emplacements ( 9 points)
- Targeting Scrambler ( 5 points)
- Grav Shift Reroute ( 2 points)
= 151 total ship cost

1 Colonel Jendon ( 20 points)
1 Maarek Steele ( 21 points)
1 Ciena Ree ( 17 points)
1 Morna Kee ( 27 points)
1 Boba Fett ( 26 points)
1 Firespray-31 ( 18 points)
= 129 total squadron cost

Hoping to diversify the stats with Screed and the Interdictor. I saw lots of variety with Sloane and Jerjerrod fairly well represented. I know Giled was gathering all the lists, but there we have it.

1 hour ago, CommanderDave said:

I took the win in Orlando with my list:

Screed Interdictor & Rogues
Author: DaveG

Faction: Galactic Empire
Points: 397/400

Commander: Admiral Screed

Assault Objective: Most Wanted
Defense Objective: Fighter Ambush
Navigation Objective: Superior Positions

Gladiator II-Class Star Destroyer (62 points)
- Demolisher ( 10 points)
- Agent Kallus ( 3 points)
- Ordnance Experts ( 4 points)
- Engine Techs ( 8 points)
- Flechette Torpedoes ( 3 points)
= 90 total ship cost

Gozanti-class Cruisers (23 points)
- Hondo Ohnaka ( 2 points)
- Comms Net ( 2 points)
= 27 total ship cost

[ flagship ] Interdictor-class Suppression Refit (90 points)
- Admiral Screed ( 26 points)
- Interdictor ( 3 points)
- Captain Brunson ( 5 points)
- Engine Techs ( 8 points)
- Disposable Capacitors ( 3 points)
- Heavy Ion Emplacements ( 9 points)
- Targeting Scrambler ( 5 points)
- Grav Shift Reroute ( 2 points)
= 151 total ship cost

1 Colonel Jendon ( 20 points)
1 Maarek Steele ( 21 points)
1 Ciena Ree ( 17 points)
1 Morna Kee ( 27 points)
1 Boba Fett ( 26 points)
1 Firespray-31 ( 18 points)
= 129 total squadron cost

Hoping to diversify the stats with Screed and the Interdictor. I saw lots of variety with Sloane and Jerjerrod fairly well represented. I know Giled was gathering all the lists, but there we have it.

Good stuff!

Hi Truthiness and other cream of the crop in Armada coverage!

So. As a budding data scientist and semi-retired intelligence officer this project looks absolutely terrific.

I'm an avid supporter of the theory that most sports/games lack a 'moneyball revolution'.

My thinking would be that data coverage in this case will always be rather thin and thus many standard tools of statistical analysis will be handicapped somewhat.

What 'original' formulaes/coefficients would best serve the role of sabermetrics in Armada?

I believe that your (Truthiness') High Performance-Frequency coefficient, as in commander frequency in top 4 vs. overall frequency in the population is one of the best ideas I've seen to describe power level of commanders. Admittedly this is one of the simpler concepts in Armada, and as my friend posits - commander power levels are confounded quite a bit by archetypes and their relative power level (obviously, we couls and should control for this factor).

Beyond simply hoping for a good discussion here, I'd like to pledge my assistance in collecting all lists from the two Regionals (to my knowledge) held here in Denmark. I hope one of them to be my first real tournament, but the entire Danish Armada community is pretty much in a small little FB group so I'm rather sure I can get them all collected either way.

Secondly, one of my gaming group members is a national champion and 2017 participant in the Excel World Cup. So if you have some super persistent issue with the spreadsheet, I'm sure I can nudge him into helping (I'm sure the first thing he'd say is to never use Google sheets for analysis though 🤣).

9 hours ago, Truthiness said:

Good stuff!

I have them all, and will get to posting them eventually. Promise. Remind me Thursday evening if you haven't heard from me on this by then.

1 hour ago, GiledPallaeon said:

I have them all, and will get to posting them eventually. Promise. Remind me Thursday evening if you haven't heard from me on this by then.

Feel free to just take pictures and send them if you like.

Any reports from Missouri or the Land Down Under?

Here, have a smile and a vegemite sandwich.

2 hours ago, Truthiness said:

Any reports from Missouri or the Land Down Under?

QLD had their first this weekend. We've got the lists but I have to find some time to type them up.

Are you trying to tempt me because I come from the land of plenty?

13 hours ago, Truthiness said:

Feel free to just take pictures and send them if you like.

Any reports from Missouri or the Land Down Under?

MO is a work in progress. The twins caught a stomach bug, and I couldn't leave the wife to deal with that alone. (What we really needed was a young priest and an old priest.)

That said, I know I can pretty easily get the top three. I don't know what else was collected.

I was at the MO regional and have all but 4 lists in a google doc. I’m going to try to get them today and will shoot you the link to the doc once I’m back at my computer.

I was at the MO regionals, I know one guy was taking photos of all the lists, and the event host was also collecting all the lists. So that should become available at some point.

7 hours ago, Belisarius09 said:

I was at the MO regionals, I know one guy was taking photos of all the lists, and the event host was also collecting all the lists. So that should become available at some point.

Yeah that was me taking the photos.

13 minutes ago, Vath said:

Yeah that was me taking the photos.

Ah, that means you were the guy who 10-1 me in round 2! Good times... Heck at least I bounced back in round 3 with a 10-1 of my own and finished 6th place. (note to self don't pick most wanted...) It also doesn't make me feel as bad knowing you got 2nd overall 😅 gg and wp, and congrats btw on getting the dice! I know that was both of our goals.

Can't you hear, can't you hear the thunder?

9 hours ago, Belisarius09 said:

Ah, that means you were the guy who 10-1 me in round 2! Good times... Heck at least I bounced back in round 3 with a 10-1 of my own and finished 6th place. (note to self don't pick most wanted...) It also doesn't make me feel as bad knowing you got 2nd overall 😅 gg and wp, and congrats btw on getting the dice! I know that was both of our goals.

ggwp to you too! Our slug fest was quite fun.

Truthiness, you should have the doc in a pm now.

Have you considered using actual statistical analysis rather than simply descriptive statistics, @Truthiness?

I believe you could develop the ratio of used in top 4 vs. used at large into an actual coefficient as said earlier. But running simple distribution analysis to see if the difference is significant rather than a chance level occurrence seems like the simplest improvement over naked eye analysis.

57 minutes ago, Cpt. Caine said:

Have you considered using actual statistical analysis rather than simply descriptive statistics, @Truthiness?

I believe you could develop the ratio of used in top 4 vs. used at large into an actual coefficient as said earlier. But running simple distribution analysis to see if the difference is significant rather than a chance level occurrence seems like the simplest improvement over naked eye analysis.

IIRC in last year's analysis there were stats showing % appearance in top 4 vs overall so you can see what appears more prominently in the top lists.

4 minutes ago, Zamalekite said:

IIRC in last year's analysis there were stats showing % appearance in top 4 vs overall so you can see what appears more prominently in the top lists.

Exactly. I think it is one of the most promising things to look at. But.

A simple percentage does little to tell you if that happened by chance or with significant probability due to 'actual' differences. What I was getting at, was improving on the metric to actually be statistically sound and developing a proper coefficient that is also, with the naked eye, comparable. I.e. develop a value of 0-1 and attach a significance analysis of distribution.

5 minutes ago, Cpt. Caine said:

Exactly. I think it is one of the most promising things to look at. But.

A simple percentage does little to tell you if that happened by chance or with significant probability due to 'actual' differences. What I was getting at, was improving on the metric to actually be statistically sound and developing a proper coefficient that is also, with the naked eye, comparable. I.e. develop a value of 0-1 and attach a significance analysis of distribution.

I think this idea is intriguing, and I actually am planning to do some exploratory analysis with the data as it is gathered. Two main issues with the idea are:

• Statistics. In reality the statistics are quite low, and the lists are likely to be biased toward the local meta. These skews are a bit harder to account for in a single coefficient without also having a comparison for it against "local coefficients".
• It takes a lot of work. Truthiness is gathering, inputting, and doing analysis. Something a little more robust may give more insight, but it is debatable on if it adds enough value to be worth the effort.

That said, I'm going to play around with it for my own benefit of buffing up some analytics skills. Feel free to do the same as the data comes around.

19 minutes ago, Astrodar said:

I think this idea is intriguing, and I actually am planning to do some exploratory analysis with the data as it is gathered. Two main issues with the idea are:

• Statistics. In reality the statistics are quite low, and the lists are likely to be biased toward the local meta. These skews are a bit harder to account for in a single coefficient without also having a comparison for it against "local coefficients".
• It takes a lot of work. Truthiness is gathering, inputting, and doing analysis. Something a little more robust may give more insight, but it is debatable on if it adds enough value to be worth the effort.

That said, I'm going to play around with it for my own benefit of buffing up some analytics skills. Feel free to do the same as the data comes around.

Just to get what you're saying. Are you saying the sample is very small? Because potentially we might be able to collect a petty big sample by my estimation, big enough for power analysis to check out

On the other hand, we might technically not be looking at ONE population (Regional fleets in Armada 2018), but MULTIPLE populations made up of a number of regionals that have similar enough metas to not differ substantially (e.g. squadron heavy meta, squadron less dominated meta). And in that case... I'm not sure what the consequence of that is, other than the ability to perhaps a more fine grained analysis of the power level of cards in the context of different metas. I mean, theoretically, we could distinguish between metas.

Thinking six steps ahead and beyond the possibilities I'd consider realistic, a modelling of all of this in the vein of the study of game theory and evolutionary stable strategies would probably be the ultimate test of objective power levels in Armada - without the problem of collecting the data