What I'd change and how I'd do it

[Jester]

This thread has been brewing for a long time. In it I'm going to try to hammer out what I'd do if I were in charge of Planetarion.

Not every idea is as well thought out and I do not claim that implementing all the changes proposed will lead to a revitalization of Planetarion. I have a good deal of assumptions about what the negative effects of the current system are, and the changes I propose are mainly changes that would challenge those assumptions. This sort of experimentation has been used extensively in Planetarion's recent development, so hopefully minds will remain open. You might say that I have a set of theories of how to improve Planetarion, and that I am willing to test them by playing in any round that implements them.

In addition to being willing to test them myself, I need to convince PAteam that they are worth implementing. From experience assisting in the development of PA, I know that the more coding work required for a feature, the less likely it is to be implemented (either promptly or at all). Therefore, each feature is examined from several angles, those that require more coding and those that require less. In addition, cohesive (and hopefully rational) arguments are provided in an attempt to convince PAteam that this experiment is worth making.

For the purposes of this thread, let 'bad' for Planetarion mean less fun and 'good' mean more fun. Let the assumption hold that a less fun game attracts fewer players and a more fun game attracts more players. Therefore, the metric of fun is playing population. If fun leads to population, and populations leads to fun, these effects are reinforcing.

Without the assumption that fun leads to population, it still holds that population leads to fun, and therefore any change that leads to a higher playing population is 'good'. This leads to the my first assumption.

Theory #1: The alliance member limit is bad.
The alliance member limit favors elite alliances and prevents mid-tier alliances from competing realistically. It removes dimensions from the alliance meta-game that were previously available.

Since the advent of the tag system, alliances have been limited to a certain number of members in various fashions. This system was put in place for many reasons, primarily to level the playing field in some fashion. In some ways it has done this, in other ways it has failed. Particularly, of the last 9 rounds, 2 alliances have won all but 1 round. These two alliances, 1up and Exilition, have both been considered the strongest and most influential during these rounds. The fact that several rounds have been won by alliances that had less members than the maximum member limit supports this.

The alliance member limit has two effects which I identify as the primary bad effects. It is difficult for players to reach the higher alliance tiers. There is little incentive for alliances to take risks on unknown players. These claims are examined and justified as bad later.

The alliance member has one effect which is commonly identifies as a good effect. It is commonly accepted as good that the winning alliance is not chosen on the sole criteria of being able to recruit most players. This claim is examined later as well.

The Market for Lemons
The market for lemons is a Nobel prize winning article about economic situations with asymmetrical information. A lemons market occurs when it is impossible for buyers to know whether they are paying for high quality or low quality (lemons), but incentive exists for sellers to pass off low quality as high.

The market for alliance members is a market for lemons. Each alliance member costs one alliance member slot, but the quality of this member will vary.

From the above linked wikipedia article, these criteria are presented for a lemon market:

1. Asymmetry of information
   • no buyers can accurately assess the value of a product through examination before sale is made
   • all sellers can more accurately assess the value of a product prior to sale
2. An incentive exists for the seller to pass off a low quality product as a higher quality one
3. Sellers have no credible disclosure technology (sellers with a great car have no way to credibly disclose this to buyers)
4. Deficiency of *effective* public quality assurances (by reputation or regulation)
5. Deficiency of *effective* guarantees / warranties

In Planetarion, these all hold for alliance members. A partial exception is made in the case of the vouching system. There is an asymmetry of information. The buyer, an alliance recruiter, can not assess the value of a member prior to the sale. The seller, the potential recruit, knows whether or not they are a good alliance member not. The potential recruit has an incentive to pass themselves off as a good member, because they are likely to receive defense from the alliance which will improve their rank. A good alliance member has no way to show that are a good one since alliance defense point systems are not open to public scrutiny. Even if they were, past records are not guarantees of future performance. While vouching systems provide public quality assurances, these are not necessarily effective, since vouchers also have incentives to lie on behalf of or at least vet their friends more strongly than they deserve. There is no way to redeem defense fleets spent on a lemon member, therefore no effective guarantees or warranties.

The alliance score calculation has contributed a slight balancing of this situation by increasing the cost of being kicked from an alliance, in the sense that it is more difficult to convince a new alliance to accept a potential recruit based on current score.

Since the alliance member market is a market for lemons, it is difficult for unknown players (new players) to join alliances that fit their skill and activity level. Since lower ranked alliances are unable to put pressure on higher ranked alliances solely by recruiting beyond the member limit, there is no incentive for higher ranked alliances to compete by recruiting beyond the member limit. That is, member numbers of alliances will stabilize below or around the current member limit. This may seem like tautology, but it is important to realize what this means. If an alliance is to be competitive, it need not concern itself with recruiting past about 80% of the alliance member limit. This means that the buyer can afford to be extremely picky once their stall is filling up. There is no incentive to take risks, and risk-seeking behavior is what catalyzes change.

The commonly accepted good consequence of the alliance member limit is that it prevents any alliance from winning based solely on recruiting. However, this is logically incoherent. If any alliance could win based on recruiting planets without any sort of cooperation, then surely an alliance could win by recruiting as aggressively while encouraging alliance cooperation? And if this is the case, then the only objection is that some people would rather win by cooperating with a small set of people than by cooperating with a large set of people.

Solutions
The solution becomes how to break the market for lemons. One way is to lower the alliance member limit so that the recruiter will have first-hand knowledge of more potential members in ratio to the number of member slots. The community has strongly objected to lowering the alliance member limit below 50 members, but even this number is too high to break the market for lemons. The objection stems from the diverging goals of alliances. Some alliances exist to compete under the rules as defined by the game system. Other alliances exist to provide social environments
in which players can play competitively. The latter group of alliances are generally those which regularly finish outside the top5 without realistic ambitions of finishing 1st.

Excluding these protests, the likely consequences are an increased competitive playing field with elite groups vying for victory. The requirements of these groups would be high levels of activity and mutual dedication. Therefore, unknown players would not be recruited into them. The value of a member slot would be too high compared to the relative risk of granting one to an unknown player.

Therefore, the solution is to remove the alliance member limit. Removing the limit would make the member slot cost of recruiting a member irrelevant, therefore encouraging alliances to take in new members with lower requirement of prior knowledge. It would encourage alliances to continue recruiting regardless of current member count. The most obvious advantage here is that members would be encouraged to market Planetarion to get more players for their alliance.

Problems
Removing the alliance member limit would encourage alliances to recruit heavily, thus creating a larger demand for members. This would hurt some alliances, and might lead to a smaller playing field.

The greatest problem is the 'recruiting to victory' problem. This problem occurs when an alliance defeats another solely by recruiting. This problem is not actually tied to the member limit, so much as it is tied to the criteria for winning. What does having the highest alliance score represent? This is covered later.

Implementation
It is clear that the alliance member limit serves no real purpose and that it would be better to remove it. This can be tested by running a round without one, or with one so high that it is in practice a non-limit. This implementation is changed almost every round and can thus be assumed to be a trivial change. The time-cost for PAteam is thus low, and the players can help show whether it is a viable option by playing the round. They can vote against the change by not playing the round (or by playing a free planet).

Theory #2: Ship stats are too complicated
The number of numbers in the ship stats is quite simply overwhelming. Most people don't look at the stats as a whole, but instead use heuristics to choose their race and use IRC bots and battlecalcs to decide targets and calculate battles. Requiring that players learn to use external tools to play the game isn't ideal, but more than that, the fundamental functionality of the game is hidden behind 55 ships with 10 variables each. 550 numbers are not something people will want to wade through to learn the game.

With the introduction of Eitraides, even the 55 ship names are too much to remember reliably. At a very minimum, a competitive player must be familiar with names, their associated ship classes, targeting and initiative. Most competitive alliances provide IRC bots to assist in this. Without access, it's impossible to play PA competitively without a battle calculator.

Tied to this is the fact that choice of race, which is done before any experience is gained with the active stats set, decides a large part of the game for players. The correct choice of race is a large factor in the enjoyment of a round. Given the difficulty of designing stats that provide equally good stats for all choices provided people choose race based on their play style preferences, it is unreasonable to expect an average player to be able to make the optimal choice.

Solutions
Simplifying stats is done by reducing the number of ships. Currently each race has 11 ships. Reducing the number of ships across the races would cut down on the depth of play, while even at a minimal 7 ships, there would still be 35 ships to memorize. This solution is not optimal.

The best solution is to make some ships common across all races. The extreme version of this is to make all ships common across the races, and instead of using race to provide ship variability, either have none or have a tech tree that decides (as was the situation prior to round 5).

Considerations should also be made to what level of changes should be expected between each round within the stats. It might be beneficial to consider limiting changes to incremental improvements or adjustments of current sets instead of reinventing the wheel each round.

What would the point of races be after removing most of the ship differences? There would still be special rules for each race. Some of them could be changed on top of ship changes. For example, consider all races having the same cloaked ships, but the rules for Xandathrii cloaking be slightly different, for example that in addition to not showing on unit scans, the Xandathrii cloaked ships don't show on galstatus or overview. This is just one example, others are possible.

The base ability differences of races also form the character of each race. The construction, research, covert operation and similar bonuses all contribute variation to the races.

Problems
Giving each race the same ships risks oversimplifying the combat options, especially with the current combat engine. The game wouldn't be much fun if every combat was exactly alike. In round 4 the situation was at its worst, not only was every combat alike, but no ships ever died.

Implementation
At worst, common ships could be implemented by having race specific versions of each ship that have different names but similar stats. The simplest form of this would be <Race> <Shipname>, for example 'Xan Interceptor' which is exactly the same as 'Ter Interceptor' except for the name. This is possible to do in the current code without alteration. If it is possible to designate ships available to multiple races, this problem does not exist.

The number of ships common should probably be at least 7, with one ship targeting each class and one common astropod. In this case, one special pod, one special Structure Killer and up to 3 special ships would be provided per race. In this case there would be 7 + 15 = 22 at the low end of the spectrum and 7 + 25 = 32 ships at the high end. While this number is only slightly lower than the 35 at 7 ships per race, it represents a much richer depth. It also represents a simplification in choosing race and coordinating with galmates, since at least some ships are common across race choice.

At the other extreme would be making all the ships entirely similar for all races. This option would reduce the number of unique ships drastically, to approximately 12, given the setup that has been common over the last rounds. It would also be possible to provide more ships under this option.

Theory #3: The scoring system is broken
It's been painfully obvious to a lot of people since round 16 (and some earlier) that the scoring system is broken. The current system of XP combined with value to determine score is based on the old system (value equates score) with a permanent reward for attacking larger planets. This was meant as a way to encourage behavior other than bashing. Whether it reduced the amount of bashing is difficult to know (it would require logs of attacks in rounds previous to and after XP was introduced).

What is certain is that scoring system as a whole has changed the priorities of players and how the best alliance is judged. A player is rewarded for making attacks that the XP formula reward. An alliance is rewarded for having players with high scores. These are two problematic situations.

To illustrate how they are incorrect, consider Appocomaster's stated goal when designing the XP formula:

<AppocoAway> I think that XP should still be a valid way of playing, but at extremes I don't think it should perhaps be quite a "round winning" option

Assume that a valid way of playing means that it should be possible to finish in the top100 by playing for XP. Define playing for XP as not sending any defense, attacking with 3 fleets every night. More proficient XP players will earn more XP than other XP players using this tactic.

The current definition of an alliance is the sum of the score of the planets in the tag. Therefore, the winning alliance is the alliance whose players total the highest score. However, to win as an alliance, one does not need to have the winning planet.

Mathematically, all one needs is an average score that would equal a rank of the maximum alliance members times 2 minus 1. (The proof here is trivial.) Therefore, Appocomaster's stated intention specifically allows alliance tags that consist of alliances made up entirely of XP players. Therefore, XP is a "round winning" option for alliances, but this violates the intention set forth.

In other words, either the ability to play for XP or the alliance tag scoring system is broken. PAteam tried to address this by changing the alliance tag scoring system to only take into account the value of the planets added to the tag. There was some backlash, and they changed it back.

Solutions
It is impossible to say which is broken without designating what behavior to reward. A good case is made for a new scoring system in this thread. In it, Heartless suggests that score should be based on spent resources rather than anything else. An amendment that would allow PAteam to remove the restrictions on galaxy and alliance donations would be that score is based on resources mined from asteroids only. It is important that it is gained only from mining asteroids, since there is already a very strong incentive to initiate few asteroids and rely on refineries to make up for this.

Ships would still have an inherent value, and it would be prudent to publish this along with score as is currently done, but it would not play any part in determining the rank of a planet. Players would be rewarded for both their skill in attacking and their ability to cooperate and defend.

The superiority of this system over the system where alliance score is designated by combined planet value is illustrated by an example. The alliance with most combined value in a round have the most ships. But conflict in Planetarion kills ships. Therefore, an alliance whose goal is high value should avoid conflict. In contrast, the scoring rule rewards an alliance for having gained a ship, regardless of whether it is later spent in combat.

This system also prevents newbies and inactives from losing what rank they have accumulated during a round, which was one of the stated goals for implementing the XP system.

Problems
The most obvious problem is that this system forces players and alliances to think ahead when planning their politics. With value based ranks, the theoretical threat of a drop or opportunity to rise always appears to be an actual possibility. With unlosable score that can't be caught with huge XP caps, players and alliances would need to plan to begin wars in time that they could catch up and dominate their opponents. This might lead to some situations in which players feel that the winner is undeserved. An example might be a top planet that lost a large portion of its fleet during the last week, or an alliance that has spent the round avoiding war losing a swift, decisive war to a more organized alliance, but having enough of a lead to win anyway.

Implementation
Removing XP would involve a fairly large number of deletions from the code. Reworking the scoring system to take score from mined roids would entail adding some code to the function that adds resources. An extra column might be needed to handle 'overflow' resources so that rounding doesn't adversely affect rank.

If desired, removing the limits on galaxy and alliance donations would require some deletions as well. Fortunately, deletions are less work than writing new code. Aside from this, the alliance code would probably not need to be touched.

Theory #4: The exile system is bad
The current galaxy/cluster setup discourages players from passing their knowledge about how to play the game to other players. Since knowledge and information are both keys to holding interest, this is bad. The source of this discouragement is the exile system.

When a galaxy is formed, a player measures his estimated requirements against what the galaxy appears to be able to provide. If the galaxy is judged to be unable to provide the requirements, the player can choose to exile or to take control of the galaxy and attempt to improve it. These options are known as exit and voice.

Exit, Voice and Loyalty
Exit, Voice, And Loyalty (more here) is a model for describing how members of an organization can respond when they perceive that the organization is demonstrating a decrease in quality or benefit to the member. They can either choose to withdraw from the relationship (exit) or they can choose to attempt to repair or improve the relationship through communication (voice). The choice is made through a cost-benefit analysis that is influenced by loyalty.

Voice, by definition, is much more useful for Planetarion. Voice provides either useful communication, the dissemination of knowledge about how Planetarion works, or conflict, which is the cornerstone of the game. Exit does not help the game in itself, except that it opens up possibilities for voice.

In Planetarion terms, exiling or self-exiling is an exit option. Currently, voice is only possible by communicating with other members of the galaxy. When this is not possible, which will only occur when the other galaxy members are entirely unresponsive, exit is the only available option.

Assume that loyalty is defined by the number of players one has befriended in a galaxy, minus the number of players one considers to be enemies. While this linear model will not cover every case, it models the most important relationships between loyalty and galaxy composition. In this case, the size of the buddypack or private portion of a galaxy is a strong influence on loyalty. In addition, friends can agree to both choose the exit option when they feel voice has failed. In this case, loyalty plays a much smaller part in the cost-benefit analysis.

The cost of exit is measured in resources and time. The benefit is measured in playability and expected final planet and galaxy rank. The cost essentially boils down to a number of days of actual play the player must sacrifice in return for further days of improved quality of play.

Assume that a player desires to be able to provide a significant contribution to their alliance's final rank. To do so, would need to provide a significant contribution to an alliance's top120 average score rank. Assuming that planet score is partially random (or rather, so chaotic that it is meaningful to call it random), define a significant contribution to be a planet score within one standard deviation of the alliance's average score. For simplicity, define this to be a top150 finish.

Since the advent of the 'planet reset' mechanism, it is possible to virtually guarantee landing in a galaxy which allows significant contribution to be made. The worst galaxy which allows this finish depends upon the number of planets per galaxy. Paradoxically, more planets per galaxy decreases the number of galaxies this would cover in the worst case, but in practice increases it. This is because the requirements to finish in the top150 are a regular amount of defense, which is more likely to be provided by a larger galaxy.

A conservative assumption is that the galaxy would need to be able to maintain a regular top30 score rank to enable a player to contribute significantly to their alliance's final rank. Given the size of galaxies in most recent rounds, this equates to approximately 400 planet slots.

The 'planet reset' option allows a player to perform unlimited self-exiles at the cost of waiting for a day for the planet reset to complete. It is therefore not unreasonable to assume that it is during the first 2-3 weeks of the round (depending on round length) always beneficial to choose exit when confronted with a galaxy that would not allow one to make a significant contribution to final alliance rank.

In Planetarion, voice is an inefficient choice because voice alone cannot provide the same guaranteed benefits of exit.

The problem of quality seeking members of galaxies choosing exit is that it leaves players who do not choose exit in an even worse situation. In essence, if no one ever chooses voice, then bad galaxies can never get better. This is witnessed by previously competitive galaxies being completely abandoned once members begin to choose the exit option.

The consequence of exit is exclusion
Since exit is the inevitable result of competitive players in uncompetitive galaxies, competitive galaxies can rely on a continual supply of competitive players to fill their empty slots. They can therefore liberally apply exit to players who do not immediately provide promise for meaningful contributions to the galaxy (by whatever measure of judgment chosen).

In other words, the exile system is just a semi-random private galaxy system that almost completely excludes inexperienced players from competitive galaxies. What's worse, they can not experiment by creating competitive galaxies of their own, because of the limitations of the buddypack system.

The cluster system
Since players continuously choose exit over voice, the cluster system no longer promotes cohesion. Since there is no guarantee that a planet will remain in the cluster, there is no incentive to provide quid pro quo defense. Not only this, but a planet arriving in a new cluster has not built up a trust relationship with the cluster members and can therefore not expect to receive defense. Since they cannot expect to receive defense, there is no incentive to send defense. Thus the cluster system does not promote building cluster alliances. On top of this, any player not in a secure position in their galaxy risks being exiled. If they are exiled, they are unlikely to be returned to their cluster, and thus have little incentive to defend their cluster alliance in the first place. In addition, the entire argument for choosing exit over voice in the galaxy applies to the cluster as well.

However, it is worth noting that it took many rounds from the benefit of cluster alliances was taken away (the removal of the incluster attack ETA bonus) to people stopped trying to form cluster alliances or even less meaningful cluster NAPs. It is worth considering that the lack of cluster alliances this round is due to a similar effect, players are not choosing the most efficient options, but as those become more clear, more players will choose them.

Solutions
The return of attacking ETA to the cluster is a good start. The next step is to encourage voice over exit in the galaxy. This can be done in several ways. PAteam's traditional solution is to make exit cost more resources. A better solution is to increase the loyalty of players to their galaxy and (if possible) to the cluster.

One way to increase loyalty is to increase the options for cooperation within a galaxy. Currently the means of cooperation are very limited. There are two primary forms. Defense cooperation and attack cooperation. Currently, attack cooperation is naturally discouraged, since it invites counter-attacks. Defense cooperation is only feasible if fleets are large enough. But since inexperienced players are unlikely to have or be able to gain large fleets, there is currently little to inspire loyalty by defense.

Allowing ingalaxy attacks and farming is one way to increase cooperation. This would provide a way for players to contribute to the whole galaxy without necessarily having to be active enough to provide defense.

Allowing ingalaxy attacks also increases the options for voice in a bad galaxy. Threatening to attack is another way to argue for an improvement in quality.

The proposed new scoring system would increase incentive to cooperate by allowing players to donate resources without losing score. Currently, donating any significant amount of resources is difficult due to the limitations in place on the galaxy and alliance fund. In addition, there is a strong disincentive to make such donations, since they negatively impact one's own rank directly.

The most powerful increase in loyalty can be achieved by allowing larger buddypacks. Buddypacks of around 10 members would allow people to choose the majority of their galaxy directly, which would breed strong group loyalty. Stronger galaxy loyalty would increase the incentives to cooperate with the cluster. Therefore, the cluster ETA changes would counter-act the lessened intermingling of social groups that larger private buddypacks would create.

Problems
Allowing farming would reduce the perceived merit of the top planets, especially if they attacked for large roid benefits in the later stages of the game. This was a major complaint in the rounds when farms were legal (r1-6). However, several limitations would remain in place to prevent this. Farming from allied planets outside the galaxy would remain impossible. Friends in other alliances might allow this, and it would be up to the leaders of these alliances to determine whether to allow this or punish it.

Ingalaxy attacks pose a greater problem in combination with the exiling system. A planet could exile to a new galaxy in the middle of the night and launch very low ETA attacks on unsuspecting targets. At the very least, a limit on the ability to attack and be attacked by a planet such as within the first 24 ticks that it is in a galaxy would be needed.

The problems associated with the new scoring system are examined above.

Under the current system, larger buddypacks would create larger galaxies. Reworking the current system so that single players are treated as buddypacks of one would help alleviate this somewhat. However, this would prevent new players (who are indistinguishable from experienced players going random) from being able to join the largest galaxies. While an improvement over the current system, this is obviously not enough for the long term. However, more advanced systems would require more coding, and must therefore be relegated future discussion.

Implementation
Removing the regulations on farming would be trivial, the multihunters would only need new instructions. Ingalaxy attacks would require removing a single check in the missions page. The changes required to support larger buddypacks would depend on how generic the code is currently.

What's right
There are a lot of things PAteam are doing right. The Planetarion interface isn't considered particularly beautiful, but it is much better than most other online browser based games. The current process of iterative improving the interface, the look and feel of PA, is healthy and should continue. PAteam have shown a strong will to continually improve Planetarion, and their iterative, experimental approach is generally working. They have also shown a strong willingness to work with the community when reasonable suggestions are made that would improve the game. Examples of this are the 'trusted/paid' tags for the Message from Your Commander, the alliance merging system, the removing of military scans and so on.

Summary
Planetarion needs change, but there is a very limited amount of energy available to implement this change. Focusing development on the main problems, and focusing it on making as much progress as possible with as little work as possible could help revitalize the game. The outlined problems are the ones I think are most fundamental to the Planetarion game engine.

I ask that any discussion in this thread be kept to the above points. If you have any further ideas, please post a new thread and link to it.