Popularity of techs

[ffm2]

Here is the popularity of the technologies in my database.

All data from A27 A24. The data is gathered from 369 games with 1436 players (not individuals [if a player is in 2 games its counted as 2 players]). From the 1436 players are the top 600 selected with the highest eco_score of spend resources by minute 13 selected and evaluated. So games in which the player booms and is likely left alone. The percentage shows how many of these players who can researched the technology actually researched the technology in that game. So if a tech is only available for han, and all the han players researched it, it would be 100%.

This could be used to balance some techs.

tech_stats.xlsx

Edited February 28 by ffm2

[real_tabasco_sauce]

38 minutes ago, ffm2 said:

All data from A24

!

This is really nice data here. Its still shocking how much ppl skip loom.

[Player of 0AD]

Whats the point of evaluating data from Alpha 24 rather than Alpha 26 and Alpha 27?

[real_tabasco_sauce]

1 hour ago, Player of 0AD said:

Whats the point of evaluating data from Alpha 24 rather than Alpha 26 and Alpha 27?

its a typo. You can see that a27 techs are in there.

[Seleucids]

Thank you for the analysis ffm !

As expected, most are eco techs, then armour, then civ-special. I see no particularly OP tech here; the most popular are all essential things needed to unlock more. 

[Emacz]

well and soemthing like helot economy is ONLY useful with one civ  Doesnt that affect thigns?

[chrstgtr]

The most interesting thing is that there are many techs that are never used.
 

Some of this might a function of a lot of these being new techs that people haven’t sufficiently tested yet (the helot tech, maybe).
 

But there are a ton of techs just never researched. Those seem like a good place to revamp or delete the techs. For example, there are a bunch of navy techs that we discussed in the other thread that just aren’t used. Other stuff like spies have been around forever and serve zero purpose. No need to overly complicate the tech tree with useless stuff. 

[ffm2]

It is data from A27 and was indeed a typo. Thanks for figuring that out.

1 hour ago, Emacz said:

well and soemthing like helot economy is ONLY useful with one civ  Doesnt that affect thigns?

Helot is only available for sparta (and 0% of the spartans did it). Costs 200 food, 200 wood. Infantry Javelineers +100% grain gather rate, but -10% ranged attack pierce damage. I see no value in this. I'd rather pay for the opposite if + and - where switched on both.

 

1 hour ago, chrstgtr said:

For example, there are a bunch of navy techs that we discussed in the other thread that just aren’t used.

I'd need to analyze naval a bit different but that would require more work. Most games are not naval games, I dislike naval so there's even fewer in my database. Naval techs should only be analyzed if a dock is build or so.

 

The top techs might seem trivial to us but could serve for new players to focus on the most important techs.

Some techs like outpost_vision are just too expensive (only 2% did it). I liked it in the past. But 300 food is just too much. You can have a additional barrack for 300 res. And if I had the choice mid game having rather 3 or 4 barracks or more vision on the outpost, I chose the additional barrack.

[ffm2]

(Han) civil_service_01: Ministers -50% training time; civil_service_02: Ministers +50% Health; both 0%. Because ministers themself are useless and only harmful as they take up population and serve little purpose. Units gather 2% more in the 40 m radius. If you  constantly have 50 units in that radius the minister would help just as much as a regular unit.

Edited February 28 by ffm2

[Gurken Khan]

15 minutes ago, ffm2 said:

ministers ... are useless

I make them and get their health upgrade; they mostly sit in their ministry and produce trickle, I try to group as much research heavy buildings around so they can pop over and earn their cost. When I'm done with research in the fortress they join my army as champs.

[wowgetoffyourcellphone]

Unlock_females (at houses) stands at 10%, lol

[real_tabasco_sauce]

52 minutes ago, wowgetoffyourcellphone said:

Unlock_females (at houses) stands at 10%, lol

Its typically off-meta in most games. 10% is pretty expected to me.

[chrstgtr]

28 minutes ago, real_tabasco_sauce said:

Its typically off-meta in most games. 10% is pretty expected to me.

It’s the most useful or least useful tech, depending on the circumstance. 

[ffm2]

I agree. unlock_females (Fertility Festival) works correctly imo. It costs 400 res (expensive). When you develop and you'd could afford it you are almost at pop max and want more fighting men. If you are at pop max. you'd delete some woman to make room anyway. It's useful to boost eco after a heavy blow and you can ensure the safety of the woman for a while.

[Arup]

4 minutes ago, ffm2 said:

It's useful to boost eco after a heavy blow

Is it tho? 30 seconds for ONE woman... not to mention the time it takes to research it XD I dont see why I shouldnt just use the cc

[sarcoma]

u have 20+ houses, so 40+ women/minute

[ffm2]

You should definitely use your CC and not let it idle. It is 30 s for one woman per house. So if you could get 30 woman in 30 s thats where it helps. There are some conditions though that need to fit as well, like having food (or getting it from mates). Or when you received a real heavy blow and need to start a new cc in min 20.

[chrstgtr]

1 hour ago, Arup said:

Is it tho? 30 seconds for ONE woman... not to mention the time it takes to research it XD I dont see why I shouldnt just use the cc

Imagine you just lost your CC and almost your entire population. You are now left with just 30 houses and a handful of men. You can now either (A) rebuild your CC, train women, slowly build barracks, and hopefully get back to max pop in 12ish  minutes or (B) research the woman house tech, build 30 women two times at the longer 30s train time, and use those women to build up your CC, barracks, and get back to max pop in 6ish minutes. 

Which one do you choose? 

@Arup, you’re coming into these discussions hot. You tell people (who have played much longer and are generally much better players than you) that their opinions are laughable. But your logic pretty quickly falls apart when tested. You should pause before mocking others’ opinions. 

[Arup]

2 hours ago, ffm2 said:

It is 30 s for one woman per house. So if you could get 30 woman in 30 s thats where it helps.

Ah yeah, makes sense for the huge time buffer then! I usually nvr let women get caught up in the destruction of cc so I usually have enough to rebuild without needing more women so I nvr thought about it like that.. thanks

1 hour ago, chrstgtr said:

You tell people (who have played much longer and are generally much better players than you) that their opinions are laughable.

Seems we have a misunderstanding XD I never meant to "mock" your opinions and in this case, ffm's idea of the situation was different than what I was imagining, thats why I say I can use the cc

[Emacz]

On 28/02/2025 at 3:50 PM, ffm2 said:

It is data from A27 and was indeed a typo. Thanks for figuring that out.

Helot is only available for sparta (and 0% of the spartans did it). Costs 200 food, 200 wood. Infantry Javelineers +100% grain gather rate, but -10% ranged attack pierce damage. I see no value in this. I'd rather pay for the opposite if + and - where switched on both

Im playing around with

"Helots +100% grain gather rate, +50% wood/metal/stone but champions take 25% longer to train."

[hyperion]

About ministers, the bonus was intended to be much more, then there was an argument made that this needs to be useless for the sake of balance. Trying the initial intended value is still something I think is worth a shot.

Another point is just because top players don't use certain techs doesn't mean low elo players can't see value in it, so the statistics probably paint a skewed picture.

[Arup]

The guy above is straight up facts! #MMGA MAKE MINISTERS GREAT AGAIN!!

[ffm2]

Marian Reforms (roman_reforms) with 11% of roman players researching this tech are a trap (most of the time).

Skrimisher cost 50 f 50 w
have a gather rate 0.5 berry 0.25 grain 1 meat 0.75 wood 0.5 stone 0.5 metal

after Marian Reforms the skirm costs 50f 50w 15m
and the gather rates are 0.24 berry 0.12 grain 0.49 meat 0.37 wood 0.24 stone 0.24 metal

A woman has gather rates
1 berry 0.5 grain 1 meat 0.7 wood 0.35 stone 0.35 metal

Marian reforms are only useful if
A) You are about to deal the finishing blow
B) Your economy can rely completely on trade

Because you men now costs metal but can gather it less efficient than woman (and woman gather it very bad in comparison to normal men already)
No other upgrade is considered here.

[real_tabasco_sauce]

On 02/03/2025 at 1:15 PM, Arup said:

The guy above is straight up facts! #MMGA MAKE MINISTERS GREAT AGAIN!!

as far as I know, they have never been even close to good.

[ffm2]

Visualizing Technology Research Timing Based on Public Opinion

For each technology, I created a diagram showing the optimal research timing based on aggregated player data.

Key Metrics and Axes

X-axis (Progress Score): Reflects the advancement of a civilization, calculated as:

score = resourcesUsed['food']
	+ resourcesUsed['wood']
	+ resourcesUsed['metal']
	+ resourcesUsed['stone']
	- unitsCapturedValue
	- unitsLostValue
	- buildingsLostValue
	- buildingsCapturedValue
	+ lootCollected
	- tributesSent
	+ tributesReceived
	+ tradeIncome

Y-axis (Progress Score by Minute 13): Indicates how efficiently a player progressed. Higher values denote better development.

Expected Trends

• Faster-developing players tend to research technologies closer to the optimal timing.
• Slower-developing players show more deviation, ideally forming a Gaussian distribution around the optimum.
• Natural barriers (e.g., resource constraints) may prevent earlier research, allowing further refinement of the optimal timing.
• This approach leverages the Wisdom of Crowds principle.

Why Progress Score Over Time?
Using time as the X-axis would skew comparisons between fast and slow players. Instead, progress score ensures fair benchmarking. E.g. a for a faster developing player it makes more sense to research p3 earlier, so the dots would lean to the left going upwards. However, for practical player reference, iso-lines mark when different player segments (e.g., top 25%, 50%) typically achieve key progress milestones. These lines are smoothed averages from 60-player samples.

Visual Enhancements

• Top Histogram: Shows the distribution of progress scores when players research the tech.
  • Solid black line: Mean research point.
  • Dotted black lines: ±1 standard deviation (68.2% is inbetween).
  • Colored solid line: Your personal mean-helping identify biases for improvement.
• Right Histogram:
  • Blue bars: Fraction of players who researched the tech at a given progress level at 13 min.
  • Gray bars: Fraction who could have researched it but didn’t.

Note: Naval techs (requiring water maps) are currently not accounted for (less than the grey bar actually have the chance to research them).

Civilization-Specific Behavior
Technologies with differing research patterns across civilizations (e.g., Ptolemaic farming phases) are split and annotated.

Data Interpretation

• Top Players: While highlighting top performers (e.g., Stockfish, Havran) is possible, overemphasizing them may not be useful. Rankings can shift slightly with minor metric adjustments (e.g., using minute 12 instead of 13).

How to Run the Analysis

1. Place the folder in your replay directory (e.g.,~/.local/share/0ad/replays).

    In A_fastest_boom_at_time_eco.py:

• zeroad_data_folder needs to point to the game. The icons, descriptions and prices are pulled.
• Set you = "ffm" (your username).
• Adjust top = 2000 (sample size) and your_top = 49 (minimum personal records included).

In B_tech_of_fastest.py:

• Configure multiprocessing (cores, top-player highlighting).
• how many cpu cores
• you player name again

Tech Categories:

• Over_70_perc: Essential techs (good for beginners).
• Over_10_under_70_perc: Situationally valuable (deeper insights).
• Under_10_perc: Rarely used
• 0_perc: Never used

Limitations & Feedback
The tool lacks robust error handling, so use caution. Exceptions for some edge-case techs are accounted for.
I welcome reports of anomalies, I hope I didn't mess up anything big.

Enjoy optimizing your boom.

top_analysis.zip

Edited 5 hours ago by ffm2