POLL: Some Balance Changes to add to VP

[filipop]

After playing with the tweaks mod I enjoyed the vast majority of changes (especially the supermarket - that one is a really good example of a fun lategame building).

The only change I struggle with is the catapult being moved to mathematics to accommodate the archer change. While I love the slinger and think the placement of the archer in calender is a great tech choice, putting the catapult in mathematics forces the early warmonger to research the majority of ancient era techs before they can attack others. Before, accessing swordsman/catapult units could be done in only a few techs, allowing AIs and players to pose a real threat very quickly. However, now an AI or player only need to research walls and they are secured for the first hundred turns.

I feel this is hurting early AI aggression and it is frustrating as a player that my tech path early on is now the same regardless of if I'm trying early warfare. There may be some blame on how strong walls are early on but I feel the previous ability to beeline iron working was a good compromise. If you are around to explain the shift of units I'd love to hear how you came to this decision as I may be missing something.

 

[pineappledan]

Spoiler Sword Line Changes :

Swordsman Line Changes:
I reduced the base CS of swordsmen to 16. This gives a little more space to vary the strength of unique swordsmen:

• Legion remains at 18CS
• Mohawk remains at 17CS
• Kris is lowered to 16CS, same as base swordsman

Swordsmen and Longswords no longer start with the Cover I promotion. Enough other units get Cover I for free; it felt repetitive.
I have replaced free Cover I with a new free promotion for the Sword line.
Professionalism
+10 HP
+10% Ranged Defense
lost on upgrade
Tercio units also get this promotion. All Tercio units have had their CS reduced by 1 across the board to compensate.
This promotions creates a continuation of the Field Works/Entrenchment/DFPs promotion line. Now, beginning in Classical, there is a unit with this anti-ranged niche promotion line in every era.

Spoiler Archer Line changes :

Adds a new ancient era unit to the Archer line: the Slinger, available from the start of the game
Archer and Composite bowman are both made 1 tech later, stronger, and more expensive
Old Archer:
Trapping Tech / 55cost
2moves / 2Range
6RCS / 4CS
Old Composite Bowman:
Mathematics Tech / 90cost
2moves / 2Range
11RCS / 11CS


New Slinger:
Agriculture Tech / 45cost
2moves / 1Range
6RCS / 6CS
New Archer:
Calendar Tech / 70cost (same as Spearman)
2moves / 2Range
9RCS / 7CS
New Composite Bowman:
Currency Tech / 110cost (10 more than Sword, 25 less than Pike)
2moves / 2Range
14RCS / 12CS

Also changes some techs, reshuffles some things without changing the actual component:
Jungle Chop moved forward to Trapping (1 tech earlier, occupies slot of old Archer)
Catapult moved to Masonry (same tech level, Walls are now a prereq to the anti-walls unit)

This change affects several civilizations with unique Archer and CBows:
Incan Slinger:
Now a Slinger replacement. Renamed to Waraq'ak
Unlocked at game start / 45
7/7 (+1/+1)
2 Moves / 2 Range (up from 1 range)
Dazed promotion (-15% CS to units hit by Waraq'ak)

Waraq'ak no longer has withdraw on melee attack​

Babylonian Bowman:
Now and Archer replacement
Unlocks at Calendar / 70
10/9 (+1/+2)
2 Moves / 2 Range
Indirect Fire

Atlatlist:
Stays a Composite Bowman replacement
Unlocked at Mathematics (1 tech earlier) / 100cost (-10)
2moves / 2Range
14RCS / 13CS (+0/+1)
Atlatl Strike (+33% vs wounded)

Spoiler small unit CS changes :

• Scout: 10CS

Spoiler England Tweak :

England

• Ship of the Line loses Indomitable promotion. Gains Splash I and II promotions.

Spoiler Songhai Tweak :

Songhai

• Removed War Canoes promotion

Spoiler Refinery Tweak :

Refinery
Moved back 1 tech to Plastics
Increased cost to 2000 (up from 1800)
Now gives 1 Coal, Iron and Oil

 

[Stalker0]

I did some statistical analysis on the Poll to see which results hold up. Because of the very small samples we have and its pretty unknown how large our "total user base is", I went with pretty tight confidence numbers (98%) and an unlimited population (which makes the variance a bit larger). The +/- variances on most of these are large (+/- 14-16% in many cases). In laymen's terms, that means that a Yes result here is pretty definitive, as even with that large variance the Yes result still won in a definite statistical way.

Note that I am using a NO result either because the poll was definitively no, or because the results were too close to be statistically relevant.

So here are the results:

Sword Line - Yes
Skirmisher Line - No
Archer Line - Yes
Scout Change - Yes
English UU - No
Spain UU - Yes
Songhai Nerf - No
Brazil Change - No
Korea Change - No
Power Plant - Yes
Oil Refinery - Yes
Tourism - Yes
Instant Tourism - Yes
Supermarket - Yes

 

[Gazebo]

I did some statistical analysis on the Poll to see which results hold up. Because of the very small samples we have and its pretty unknown how large our "total user base is", I went with pretty tight confidence numbers (98%) and an unlimited population (which makes the variance a bit larger). The +/- variances on most of these are large (+/- 14-16% in many cases). In laymen's terms, that means that a Yes result here is pretty definitive, as even with that large variance the Yes result still won in a definite statistical way.

Note that I am using a NO result either because the poll was definitively no, or because the results were too close to be statistically relevant.

So here are the results:

Sword Line - Yes
Skirmisher Line - No
Archer Line - Yes
Scout Change - Yes
English UU - No
Spain UU - Yes
Songhai Nerf - No
Brazil Change - No
Korea Change - No
Power Plant - Yes
Oil Refinery - Yes
Tourism - No
Instant Tourism - Yes
Supermarket - Yes

Unless we get a massive influx of votes (unlikely), I'm fine with moving on the 'YES' votes here. @pineappledan you may have better luck on a referendum on the others if, after implementation, you run a second scaled-down poll.

 

[Rekk]

I did some statistical analysis on the Poll to see which results hold up. Because of the very small samples we have and its pretty unknown how large our "total user base is", I went with pretty tight confidence numbers (98%) and an unlimited population (which makes the variance a bit larger). The +/- variances on most of these are large (+/- 14-16% in many cases). In laymen's terms, that means that a Yes result here is pretty definitive, as even with that large variance the Yes result still won in a definite statistical way.

Note that I am using a NO result either because the poll was definitively no, or because the results were too close to be statistically relevant.

How does your analysis handle abstainers? ie. there are 81 total voters, but most of the polls don't add up to 81 votes cast.

 

[Stalker0]

How does your analysis handle abstainers? ie. there are 81 total voters, but most of the polls don't add up to 81 votes cast.

I looked at each yes/no pair.

So for example: if one pair was 50 yes, 25 no…that’s 75 votes for the purpose of margin of error calculation.

It’s the reason I complemented this format when it first came out, because it lets me handle these with simple statistics, whereas with normal multichoice polls I don’t have the tools to handle them well

 

[Stalker0]

In the interest of transparency, let me walk through an example of how my numbers are calculated:

The goal of this calculation is to ask: Is the Yes result in a poll "definitive" and truly a representation of what people think....or is it just a fluke that results because so few people responded.

Lets take the first pair of numbers: The sword line

Yes: 47
No: 20
Total: 67

Yes %: 47/ 67 = 70.15%
No % 20 / 67 = 29.85%

Now we calculate our margin of errors: I use a standard sample margin calculator like the one found here: https://www.calculator.net/sample-s...=2&cl2=98&ss2=67&pc2=50&ps2=&x=65&y=17#findci

Confidence Level: 98%
Sample Size: 67 (aka our total votes for the pair)
Population Proportion: 50% (not really a factor in our work here)
Population Size: Unlimited

In this example the margin is: +/- 14.23% (yes the margins are large, this is the price you pay for having such a low number of votes).

We now apply our margins to our numbers. We subtract the margin from the higher number, and add it to the lower number. This represents the maximum variation that could have occurred in our results (in other words the biggest difference in result as compared to the base result that is still just as statistically likely).

70.15 - 14.23 = 55.92%
29.85 + 14.23 = 44.08%

so what this says is: a 55.92/44.08 split is just as likely our result as the 70.15/29.85 initial result.....or anywhere in between (60/40, 65/35....all equally likely, etc).


Finally: We check for an inversion. If the higher number remains higher after our margin of error (which it does in this example), then the result is definitive. This means we can statistically say there is a preference based on the data (in our closest race, the YES result still wins out). If the numbers flip, and the lower number actually becomes the highest after we apply the error....than that means we cannot say anything about the results (could have been yes, could have been no, could have been dead even....we just don't know with any confidence what the true result is).

 

[balparmak]

Shouldn't the Tourism rework fall within our confidence levels? Margin is %14.80 for n = 62 (50 vs 12 currently), which translates to 80.6 (65.8)% Yes and 19.4(34.2)% No

 

[Stalker0]

Shouldn't the Tourism rework fall within our confidence levels? Margin is %14.80 for n = 62 (50 vs 12 currently), which translates to 80.6 (65.8)% Yes and 19.4(34.2)% No

You are absolutely right, those are the numbers on my sheet, and I just plain missed it. My great apologies.