This is the final article in my Lists and Logic trilogy (See Part 1: Deduction and Part 2: Induction). In this article I discuss Abductive logic, which often considered a bastard child of Deduction and Induction. (For indepth discussions see here and here) The reason why Abductive arguments are at times treated with scorn, is that they seem to make an even bigger jump than inductive arguments: inductive arguments generalize, while abductive arguments say that successful predictions ‘prove’ a theory is true. Abductive arguments are not deductively valid, because false theories can make true predictions, so true predictions do not guarantee that the theory is true.
I want you to consider the image above. Whereas Deduction guarantees a reliable outcome if you follow its rules and Induction implies that based on experience one outcome is probably more valid than others, Abduction acknowledges that there several possible outcomes and each of them have different levels of probability but that you can know which one is the best.
It is this potential for several “true” outcomes that leads me to include Abduction in the context of Warmachine List Building. In particular, Abduction can be seen as the logic which best fits the maxim “ Lists don’t win, Players win.” Its especially important to recognize the role of the player in the execution of any list.
If there is a strong critique of applying either Deductive or Inductive Logic to the process of building a list in Warmachine/Hordes, it is that these processes attempt to achieve total reliability (Deduction) or total validity (induction) but are often unable to survive the first dice roll. Both Deduction and Induction operate under an assumption that the Player is in possession of total information and as such can lead to blind spots. If you’ve ever said after a loss, “I didn’t know X could do Y” you’ve encountered a blind spot (I will address these in a later column).
In contrast, the strength of Abductive reasoning is that it is flexible enough that you can still apply it even when you only have access to imperfect information.
Abduction: Explaining the Unexpected
At the 2013 Warmachine Weekend Invitational, our proverbial world was shaken when both players in the finals were using models that were (and still are) generally seen as less than optimal pieces. Yet the reality is that for 90+% of Warmachine players, this hasn’t suddenly changed their conceptualization of list building. To put this in terms of List Building, most people are typically familiar with deductive structures to arguments. Below is a deductive argument that might have gone through many people’s heads when they heard that Pagani used Assault Kommandos and Flanzer used Sons of Bragg and Dhalia & Skarrath
P: Good Lists are built with Good models
P: Pagani/Flanzer built Goods lists because they wanted to do well at Warmachine Weekend
C: Since Pagani/Flanzer did well, the models they used are good
However, what deductive logic fails to do is to take account of several external factors. Deduction based on observation is on the outside looking in. This means that it is unable to incorporate useful revisions and and often fails to appreciate the ability of players like Pagani and Flanzer to make the best use of models that have specific strengths. Below is the structure of an Abductive argument.
- The surprising fact, C, is observed.
- But if A/B were true, C would be a matter of course.
- Hence, there is reason to suspect that A/B is true.
When you read Will’s reasons for his inclusion of Assault Kommandos, (here) the following abductive statement is a more reliable way to understand why Will built his list the way he did:
The surprising fact, Assault Kommandos in a Invitational List [C], is observed.
But if Assault Kommandos are good against Cryx/Legion/Menoth [A] AND high probability of several Cryx/Legion/Menoth opponents [B] were true, [C] would be a matter of course.
Hence, there is reason to suspect that [A] AND [B] is true.
It is this reliance on Inference to the Best Explanation which the biggest difference between a deductive truth and an abductive probability. The conclusion of an abductive argument, the supposed “best explanation,” is not as secure as the conclusion of a sound deductive argument. It is not necessarily even probable, as is the conclusion of a strong inductive argument. Rather, the conclusion of a good abductive argument is merely the best explanation we know of.
To this end, I conclude this article with the two aspects of Abductive Logic that help build your list assuming you are aware about what you don’t know. The first is to use Abductive logic as Revisionary, used to on modify beliefs and change decisions made based on the available evidence. The second is to use Abductive logic as Ampliative, to justify decisions based on assumptions that aren’t always supported by the premises.
Abduction as Revision: Sometimes You’re Completely Wrong (about the little things)
As way to improve your list building, it is important to note that all Players make Revisions of what is good and bad happen all the time. This the part of Abduction that feels the most like Deduction. Sometimes these are small tactical revisions such as “I need to make sure I deploy my army better.” Other times these are large strategic revisions “Its probable that I should never drop Haley Double Stormwall against Skorne.”
Yet as I’ve noted before, when models like the Mountain King are a part of lists which win Hardcore or when we look at results such as Flanzer v Pagani, in which both players played models that are often sneered at by the larger community it is possible to declare “Everything I thought I knew is worthless!” Yet the this would be rather radical response to limited data.
Why? Because generally, it is wisest to follow the Bayesian approach to belief revision, which generally holds, “When changing beliefs in response to new evidence, you should continue to believe as many of the old beliefs as possible”.
If prior to Will’s win at Warmachine Weekend you discouraged people from using Assault Kommandos, you weren’t wrong about the “big” point: that Assault Kommandos don’t belong in most lists. However, what Will has demonstrated is that often its the small nuances of the both the lists (the fact that AKs are in Will’s Vlad1 lists and that Signs and Portents certainly helps should not go unnoticed) and the matchups in which he used them.
The lesson here is to remember that though you may be right about a model, a list, or a matchup 90% percent of the time, there remains the 10% of the time where it is important to allow that a model you have discounted before will be useful.
Ampliative: Sometimes your best guess is the best guess
Have you ever made a list and said “Well if I can avoid Faction/Caster X, I’m fine” or “If I see Faction/Caster Y, I’ve made the perfect list to beat them.” If so, then you’ve made an ampliative statement about your chances for success. Will’s choice to include Assault Kommandos and Jason’s choice to include the Sons of Bragg and Dhalia and Skarrath are prime examples of Ampliative reasoning.
I say this not to insinuate that Flanzer and Pagani were lucky but rather that they were able to employ pieces that have uses to provide them tools that they might not have otherwise. What is important about abduction compared to the other forms of logic is that players like Flanzer and Will aren’t attempting to play meta proven lists (deduction) nor are they attempting to build lists strictly based on their own experiences (induction) rather they are utilizing a mixture of the two to generate a hypothesis about what will be a good list.
What makes the use of Abduction difficult for many players, Veteran and New players a like, is that Ampliative reasoning is logically invalid. It is entirely possible that the conclusions of an ampliative argument can be false although all of its premises may be true. Consequently, the rules involved in ampliative reasoning do not guarantee that whenever the premises of an argument are true the conclusion will also be true. The reason is that as skilled as both Jason and WIll are, even they (likely) wouldn’t be so bold as to have guaranteed their success before their games nor likely would they guarantee a repeat performance.
If you recall from previous articles, I am writing this particular series of articles to encourage you to be more empirical in your list building. As such, I want to emphasize that developing the ability to use Abductive logic requires training yourself to A) recognize the strengths of 10% models and B) know when to use a 10% percent model and will take similar processes I laid out when discussing Deduction and Induction.
Yet knowing when to put marginal models into lists is only half of the abductive process. This is not a magic pathway to playing like Pagani/Flanzer. Rather the Abductive process requires that when building lists you must be comfortable making lists without over thinking them, accepting the potential for bad matchups and be willing to play the game as it lays on the table, not the game that lists appear to dictate.