Levis/weasels changes

[Shaddowknight]

Ya, levi's have been nerfed quite a bit lately, but that doesn't mean they are balanced properly. It's hard to find the perfect balance though, but what we have now seems to be okay for the time being.

[Ephemeral]

Another aspect in the consideration of the power of a ship is 'attaching'. Obviously any LT ends up with combined firepower, the data above does not discriminate in which kills may have been part of an LT.

I still maintain that the biggest issue we have is a need to curb private freq. LTs which are not trying to capture the flag room. Simply raining bombs down on two hapless teams battling it out in the flag room serves no strategic purpose, goes counter to the mission statement of the zone, and often kills good competitive basing games.
eph

[Drosophila]

Hi but..why..per..,
In reply to your assertion that the levi has never been weaker, here are the ship stats from Aug 6th, 2013. They were collected in the same way as described in the previous post.

I believe this was when levi's bombs were reduced and had emp. Perhaps someone could dig through forums for the exact differences.

As you can see the win loss ratio was 2.4, as opposed to 3.3 as it is now.

Best,
Drosophila

	ave win/loss	median win/loss	standard dev	n	tinv	std error	95% ci
shark	0.71	0.63	0.38	100	1.98	0.04	0.08
weasel	0.90	0.83	0.44	100	1.98	0.04	0.09
jav	1.00	0.92	0.41	100	1.98	0.04	0.08
spid	1.07	1.05	0.30	100	1.98	0.03	0.06
lanc	1.16	1.03	0.50	100	1.98	0.05	0.10
terr	1.20	1.10	0.47	100	1.98	0.05	0.09
wb	1.28	1.15	0.70	100	1.98	0.07	0.14
levi	2.40	2.01	1.58	100	1.98	0.16	0.31

[Drosophila]

Dro,
Thank you for taking the time to do this analysis; but I have a few concerns.

I understand why you limited it the top 100 but doesn’t this skew the results? For example it may, or may not, have captured new players. I suggest that new player stats are also as important, perhaps more important, than those of experienced players (which are more likely to be included in the Top 100) when thinking about the power of the ships.

Additionally ‘kills’ are only a small part of the basing strategy. For example a person could effectively play shark and contribute to his team without racking up many kills at all; repping the enemy mines and taking down the opposing terrier without ever laying a mine himself. A terrier could play a very strong role in a basing win by simply staying alive. My point is that simply considering ‘kills’ presents an incomplete picture of a ship strengths.

But again thank you for taking the initiative to introduce empirical data into this forum and improvement discussions. It is far better than to have this kind of information as opposed to simply personal opinions.
eph

I understand your concerns with skewing the data. By limiting it to the top 100, the data summarize the most active players. Your objection is based on the assumption that new players are less active than more experienced players. With regards to activity in pub, I'm not sure if this is correct. I say this because there is no correlation (i.e. R squared of 0.002) between the win loss ratio and the total number of wins. (This is to say if you plot out the win/loss ratio vs the total number of kills it makes a line flatter than Kansas). Also in terms of the levis, the least active player had 59 wins and 30 losses.

I also agree with your points about win/loss ratio not perfectly capturing the effectiveness of players during basing. Again the issue is not with the ships that are focused on basing, but rather with the levi being overpowered as compared to the other ships.

P.S. As a side note, I have looked at the relationship between the number of win/losses and the number of hours played for warbirds. I am glad to post this if you are interested.

[Ephemeral]

... Again the issue is not with the ships that are focused on basing, but rather with the levi being overpowered as compared to the other ships.....

I concur, it is a subset of Pub; this is why it is of limited value. Considering only weapons is a very incomplete perspective. The way the ships fly, their defenses, LTs, team killing, flag touches, and even other arena activities (i.e. kill-a-thons, dueling, betting) are all things that comprise a much more complex picture than just weapons and kill ratios.

Teamwork and capturing the flag room is supposed to be what Pub 0 is all about (as per the mission statement on the web site). Rather then the 'bottom up' approach of pulling weapon data why not use a 'top down' approach? The first thing we all learned in Stat class was that you can always find some metric that supports a specific position that you might want to take. IMO the best 'top down' place to start would be to figure out how to use empirical data for questions such as;

1. How close is Pub 0 getting to its mission statement? (What percentage of players are trying to capture flag vs dueling vs just flying around killing? etc etc)
2. How is the zone doing on new player retention? (How many player download, enter Pub 0, and then leave to never come back? etc etc)
3. How many competitive basing games are occuring? (What is the impact of other zone events on Pub 0 games? What impact does teamkilling have on competitive games? What impact do private frequency LTs have on competitive basing games? etc etc.)

Obviously the list above only touches on some of the really good data that we could be trying to gather and trend. TW, and Pub 0 itself, are far more complex than just ship weapons and demands a holistic perspective. And frankly making 'bottom up' decisions is exactly what has gotten the zone in trouble. We need to have comprehension of the all the parts of Pub 0 and how they are intimately interconnected to make good decisions.

I applaud your moving the discussion into the realm of real data, you have indeed provided one part of the bigger picture. But IMO extrapolating only this data into making decisions on changing ship configurations in something we would not want to do.
eph

[bkgmjo]

I think it would be interesting to discuss the idea of reducing the strength of the levi to bring it to a similar level as to the other ships.

Make ships even? Give all ships the same configuration.

Maybe you should think this little math game over and get the stats for 1 vs. 1 fights and do this for a duel type zone. Otherwise you are completely neglecting the fact that in TW every ship has a teaming role and special purposes.

Those numbers are nothing more than plain nonsense. A crappy attempt of removing ships you cannot handle and pushing a play style that has nothing at all to do with the zone.

[project dragon]

Some thoughts:

I think having tangible data is great, it helps us look at things with a clearer perspective. One thing though, while I'm observing this conversation, is that we might be forgetting that there are currently two types on Levi's. One is the base levi, and the other is the LT. I would argue that the LT levis would skew the w/l avg, since they can rack up the kills easily per death. But what about a base levi? There is usually more risk to the levi in base, since it has limited escape paths, and the added dangers of being teamkilled. So I wonder: which levi are we considering overpowered? Underpowered?

For me, the base levi adds an interesting dynamic into a basing game, especially in the larger bases (it's a little difficult in small base, when there is only one entrance). The goal for the team is a little bit more common: kill the levi first, maybe use more sharks to do so. It forces you to work more as a team to survive. The team with the levi also has a huge advantage in base, since it does turn into a killing machine, though there is always the risk of the levi teamkilling with a bad shot or a good nme rep. It's this ebb and flow of the base levi that I feel makes it an okay ship at the moment.

If the LT is what a lot of people are concerned about, how can we make effective change that could only affect a LT, but leave the base levi a viable option? Are there ways to change terr/levi stats once they are attached (eg: some zones reduce movement stats when someone is attached to you)? If LT's are a strategy that does not fit with the mission statement as eph points out, we should focus on ways to address that. I would say that the base levi would fit into the mission statement, so there should be ways to somehow minimize their nerfing.

[fiS]

I think the numbers are OK to use for some reference and it's a nice idea to look at a ship's average. But yes, the shark isn't supposed to go positive. (Well, of course you could be streaking as a shark, but usually one doesn't do that in a basing game but a shark is always needed or greatly appreciated because of its repels). So it's fine.

Now the levi: it actually surprised me that the average from a lev is 3:1 but the logic behind is that once a levi is greened and allowed to either LT or sit in FR protecting the flag its able to get pretty high streaks, that would explain it. You must consider though that a levi (sometimes) has a hard time to travel up to fr or to green at all, so levi's aren't constantly streaking. I hope I make sense. I would also say that an average shark gains more kills per hour (or any time frame), than a levi, and a warbird or spider would gain much more than that.

Actually I'm not sure on that, but I would imagine it to be like that. Would be interesting to verify that. But if you get, say, 10 kills in 2 minutes as WB, and 10 kills in 5 minutes in levi on average, but the levi has the 3:1 K/D, then I'd still call it balanced (just based on the numbers, taking no other matters into that)

I hope I make sense, and I think if we take a WHOLE average of pub that maybe can bring some more light rather than just the top 100, although it's just number-talk lol - interpreting these numbers is maybe fun though

[project dragon]

Also, do the pub stats take into account teamkills, or are they ignored like they are in twd (teamkills gets their own stats and are left out of the kill stats)?

[bkgmjo]

People complain about levs for ages now, yet whenever a mislead pub leader changed them population drastically dropped, so why not move this kind of conversations into trash talk instead of hiding actual cheating by staff members there?

[Drosophila]

Hi Eph,
I wonder if there may be some misunderstanding. The point I am making is that the win/loss ratio of the levi is much greater than the other ships. Your point seems to be that I am 'cherry picking' the data. If so, what data would you like to see with regards to the win the loss data for the levi and other ships? I'm glad to work with you on this.

On a related note, below is a graph that shows the win loss ratio of the top 100 most active players for each of the eight ships.

As should be clear, 7 of the ships are tightly clustered together.

The levi on the other hand …...

Best,
Drosophila

[Drosophila]

I agree the questions about new player retention and considering the system as a whole are important. Closely related to this is the question of how we are doing as a zone overall.

The population stats for our zone can be viewed at:
http://sshq.net/zone/6/SSCU_Trench_Wars

I have copy pasted the stats for the past 3 1/2 years below.

In addition, I have put the initials for the months Jan, April, July and Oct above the top to make it easier to read.

The first obvious trend is that over the past 3 years there has been a decline. In July 2010 it looks like the total population was about 220. Now it is roughly 150.

In addition, there seems to be seasonal variation. Of note there is typically a peak around January and a dip between july to october.




To make it easier to look at the year to year variation, I copied the red line, turned it purple and shifted it back a year. I think this also makes it easier to see the seasonal variation.



I think it would be useful to look over various periods of time to identify why the population increased or dropped. Here are a couple of steps that would aid this.

1. Develop a timeline of who was head of pub and was in other important positions.
2. Talk to them about what was going on at the time.
3. Identify when significant changes were made to the ships, the pub map and when the buy system was created.
4. Look at how these correlate with changes in the total population.

Time periods of note seem to be:
In Oct 2010 there was a sharp drop in population.
From Oct 2010 to March 2012 things look relatively stable.
In March 2012 there was a sharp drop and then a decline that continued to late July.
The population then rose nicely to Feb 2013. Then dropped sharply until April 2013.
From April to Aug 2013 things were stable.
In August 2013 the population dropped sharply.
For Oct 2013 to Dec 2013 it rose, but not as sharply as the year before.

By doing this I think a couple of questions can be answered.
1. Are things better left alone? More precisely, do rises and declines correlate with periods when changes were made or when things where left alone.
2. Do changes in the ships have an impact on population?
3. Do changes in the maps have an impact on population?
4. Did the buy system have an impact on pop?

If you are interested in working on this with me, let me know.

Looking forward, I would suggest that if any changes are made to pub, leagues, or other aspects of TW that one change is made at a time. Likewise, that metrics are monitored before and after the change.

Best wishes,
Drosophila

[Shaddowknight]

I find it interesting that we had a "stable" period back when we had someone in charge of pub who didn't play pub or was active in any way. (If I am recalling correctly when pS was still in charge of it)

[Ephemeral]

The seasonal fluctuation is well known. I assume it is generated upon the player population being mostly located in the northern hemisphere and the likelihood that players will be inside more during winter months.

I am not sure that trying to make sense of these ‘30,000 foot’ population stats is going to yield anything meaningful. First, we really need to access to the data itself so we can query it ourselves. The pretty pictures are nice but if we had the data we do thinks such as adding trendlines and asking additional questions of the data. We currently have no way to analysis/filter the data based the most basic things such as events running or server crashes; never mind the more subtle changes such as map or ship changes.

But second and much more importantly, there are simply far too many variables to ever be able to make sense of the historical data. Forget that we can’t really account for external forces at work (i.e. some new game is released for PS4). But the real rub here is that you have multiple people making multiple changes to multiple aspects of the zone contemporaneously. While some of this may be able to be reconstructed in a timeline there are many, many things that simply are done under the radar.

Anyone who has any experience in product testing understand the importance of maintaining a stable version and controlling as many variables as possible. Having a ‘moving target’ is virtually useless in terms of analyzing anything. I would not recommend spending time on any of the historical data. IMO this is what would need to be done to get an accurate understanding of the impact to making changes to the zone.

1. Remove everyone’s ability to make any changes to anything. This goes for maps, ships, events, rules, policies, procedures, server settings, etc.
2. Downsizing some of the current features and removing some of the complexity would go a long way to help making analysis more straight forward. For example, look over the Event list and settle on the XX amount of events that should become part of the core functionality of the zone.
3. Cease the ‘gee if you have a good idea let us know’ bullshit. To some degree this is exactly what got the zone so screwed up. Too many ideas that got implemented half-assed, without being completely thought though, and with little justification to their possible impact to other areas of the zone.
4. Allow zone to settle for at least 60-90 days to start building a baseline to work with. Use this time to develop a testing plan and tools needed to gather and analyze the data.
5. Beyond a test plan we would also want to use this time to develop a real plan for the entire zone. List out the core things that exist but are broken and need to be fixed; add priorities to the list. Get scope of work estimates from those who would actually be doing the work. Put together a schedule.
6. Since there will be many, many smaller things that need to be done it only makes sense to organize them into ‘revision releases’. Intelligent grouping of these changes/improvements would go a long way to facilitating a faster testing cycle.
7. Once the baseline and plans are in place then, and only then, implement the highest priority revision release. Gather data for XX amount of time and analyze the results. Consider getting feedback from players. If all looks good move on to implementing the next release revision. If there are problems, we would need to figure out if we roll back the entire release and reanalyze or if just parts if it can be done.

If the zone had been run this way years ago we would not be in the cluster fuck we are now. Some may argue that this approach takes too much time. I would argue that it false economy to think that doing it half-assed saved any time at all (never mind the risk that was taken).
eph

[Azov]

I'd be curious to see population statistics over a longer period if they were available.