Run Away!

Breakpoints, morale failure and success in battle

As part of the forthcoming essay on the rules of infantry combat I was pondering why some units break and some don't. I started to look at the research, and as these things do, what was meant to be a couple of paragraphs in another essay turned into an essay itself. Although I started being interested in morale failure, and what was most likely to cause it, I found myself looking at what made units successful as well. I reasoned that this was the antithesis of failure so the causes of success would be the same as those for defeat but the "other way around".

What I wanted to do was to put numerical values on causes of failure and success as this is the first stage in working out some wargames rules to model the effects. Janice Fain and her co-authors in Forced Changes of Combat Posture had tried a similar approach but did not apply ratings to the elements they discovered as significant.

In this essay a "variable" is a factor that causes good or poor combat results, such as the Morale of Friendly Troops

What's a Breakpoint?

Although there are multiple definitions and some sources give more than one type of breakpoint, the one used in this essay is the point at which a unit can no longer follow its orders. Note this probably means different things depending on whether a unit is attacking or defending. An attacking unit will probably halt, or may pull back to the start line. Defending units that are defeated will retreat and this retreat may be precipitous and disordered (i.e. a rout).

Not all stops will be breakpoints, some may be temporary, for example suppression and neutralisation. Even a retreat can be stemmed by resolute leadership.

There is no attempt in this essay to quantify the magnitude of the failure and/or success and, excepting Dorothy Clark's work, and to some extent that by David Rowland, no attempt has been in the research looked at defining different levels of success or failure.

For the purposes of game mechanics we are interested in anything psychological that prevents a unit or soldier from completing their assigned mission. Given that fighting to the last man is almost unheard of its the author's contention that all defeats are psychological and morale based.

The Research

I accessed the following papers that deal with the reasons for breaking or defeat. Each one produced  dataset which fed into the overall analysis.

Battle Outcomes: Casualty Rates As a Measure of Defeat, by Robert McQuie. I only had access to a brief summary reported at: http://www.dupuyinstitute.org/blog/2017/06/23/battle-outcomes-casualty-rates-as-a-measure-of-defeat/. This work probably deals with divisional size units. It already gives percentage weighting to the variables McQuie thought important.

Forced Changes of Combat Posture, by Janice Fain et al. www.dtic.mil/dtic/tr/fulltext/u2/a201405.pdf  This reference has a very useful survey of combat veterans in it. They make personal ratings as to the the effect of each variable on the likelihood of a unit breaking. These were converted by this author into metrics by rating decisive effects twice as much as significant effects. Its worth bearing in mind that this survey had a small number of respondents, 8 discussion groups with 36 veterans in total. The veterans varied in rank and give a range of views on the performance of units estimated from platoon to regimental size. All the results of the interviews are lumped together. The effect of a respondents' rank is not further explored.

Casualties as a Measure of the Loss of Combat Effectiveness of an Infantry Battalion by Dorothy Clark is a study of the historical record for 44 US infantry battalions. www.dtic.mil/dtic/tr/fulltext/u2/059384.pdf  Clark lists the variables she thinks are pertinent to a unit not being able to complete its mission. She also makes a judgement as to which she thinks are most significant. When calculating percentages for analysis all variables were weighted the same except for those Clark judged significant which were given a double rating.

Whilst researching for this essay the historian and games designer Jack Radey and I had a private email conversation about unit collapse. Jack's replies were very insightful. It was useful to get an Eastern Font viewpoint as the majority of the data used in this essay is definitely US-centric. In order to get numeric weightings for this research, I did a content analysis of Jack's replies. This is probably the least scientific way of gathering data, so the percentages have to be treated with caution. Some variable weightings in this data are grossly out of line with all the other data. Its is for the reader to decide if that's down to poor scientific method or down to the nature of Eastern Front combat.

I accessed the following research that dealt with success on the battlefield

Understanding Defeat: How To Recover From Loss In Battle To Gain Victory In War by Trevor N. Dupuy. Unfortunately I do not have a copy of this book but there is a nice list of the factors he considered important here: http://community.battlefront.com/topic/120750-tactics-reading-undestanding-defeat-defence-of-duffers-drift-battle-of-boobys-bluffs/ . In order to convert these into numeric weightings I gave the major headings a weight of two (the numbered elements of the list) and the subheadings a weight of one (the lettered elements of the list).

Manoeuvre Warfare, some conditions associated with success at the operational level by David Rowland. http://www.dtic.mil/dtic/tr/fulltext/u2/a323544.pdf gives his top 10 variables for a successful breakthrough (i.e. an attack that penetrates the enemy's defensive line). He also gives factors for campaign success but as I'm interested in battles I only coded the breakthrough data. I used the chi-squared ratings as weightings (chi-squared is a measure of how much a variable effects the outcome). It's very interesting that although Rowland analysed many measures to do with enemy performance not a single one made the top 10 variables influencing success, all the variables in the top 10 were related to the attacking unit. There is probably something to be said about the strength of holding the initiative and whether defence is stronger that attack, but that's outside the cope of this essay. One of the top 10 variables is shock. I used the David's book, The Stress of Battle, to break down the causes of shock. Unfortunately the causes of shock overlap with each other which made it impossible to extract them all. As I already had some results by that point I chose the variables other studies had indicated were most important. However, the weightings of these variables should be regarded as tentative at best.

Richard Haynes et al wrote: Measurement of Unit Effectiveness in Marine Corps Infantry Battalions. http://www.dtic.mil/dtic/tr/fulltext/u2/a049231.pdf This is a report on several pieces of research in this area. The most suitable for our needs is chapter 6 which enumerates the chances of success in battle when a particular variable is present or not present. The metric used to convert the data into a percentage is the difference between the values when present or not present. In cases where a variable that was positive to the marines was shown to have a negative effect, and there was no readily apparent reason for this, the variable was ignored and not coded.

Richard's work frequently refers to the "principles of war" as a variable. These are not broken down further. However in order to make the data compatible with the other studies a breakdown was required. Wikipedia supplies a list of US principles of war: https://en.wikipedia.org/wiki/Principles_of_war#United_States_principles_of_war. As the author had no info on their relative importance they were all coded with equal weighting.

Finally I accesses a book that didn't reference success or failure as such, but in some ways addresses both.

Fear in Battle by John Dollard: https://babel.hathitrust.org/cgi/pt?id=mdp.39015002371410;view=1up;seq=17 is a small book that details post combat interviews with around 300 US veterans who fought in the Spanish Civil War. The work details what are the most frightening conditions and what works to combat fear, so is certainly suited to our investigations. The results are highly personal so apply only at the small unit level. As all the soldiers were volunteers who went overseas to fight against fascism this survey shows "belief in war aims" as a very strong motivational factor. Also as SCW was largely an infantry war, and the tanks used by the Nationalist were pretty pathetic, fear of tanks is notable by its absence in the survey results.

I also accessed research that only looked at what variables effected unit and soldier performance. This research took no account of possible enemy actions. It became apparent during the course of the calculations that this data had too few data points compared to the other studies so I removed them from the main database.

Method

As discussed above all the variables in each dataset were reduced to metrics. Each metric was expressed as a percentage.

All the datasets used different variables so an effort had to be made to relate them to a common set of core variables. This was in a iterative process and the core variables were updated after each dataset was added if a study contained a variable not already related to those in the core set. One core principle was to have core variables that were monolithic. For example, many of the datasets broke friendly fire support down into air and artillery. However, some did not. In order to reduce the number of arbitrary divisions the core variable selected was the top level one, in this case friendly fire support (actually coded in the table below as "firepower own".) The individual variables, and what they encompass is discussed below.

Once the core variables were selected it was a judgemental process as to which dataset variables applied to which core variables. In many cases multiple dataset variables applied to the core variable. For example recce, intelligence and surprise dataset variables were amalgamated into the surprise core variable.

The final step was to break the data down in different ways to examine the effects of the collection process on the data, and to validate amalgamating all the datasets together to produce an overall rating against each variable. For example I examined whether the datasets for divisional data were different to those for smaller units.

Core Variables

Where the variable is labelled "both" the "sense" of the variable will change depending on whether the particular dataset deals with defeat criteria or success criteria. For example, for studies dealing with defeat the manoeuvre variable would rate the importance of enemy manoeuvre. For studies dealing with success studies rate the importance of friendly manoeuvre.

Results

The data starts with the individual data from each of the data sets. These are followed by the overall ratings for different sets of data. The green cells are the highest numbers and the red cells the lowest. These are colour coded on a per column basis.

The offensive, defensive or both row lists whether the dataset takes the point of view of the attacker or defender. Note the importance of the Radey dataset as its the only one purely from the defender's point of view.

Defeat or success specifies whether the dataset is looking for variables contributing to defeat or variables contributing to success.

Nation is the primary nation represented in the dataset. The US bias is greater than it first appears as both McQuie's and Dupuy's datasets almost certainly have a preponderance of US examples.

Unit size is is the rough size of the unit(s) covered. Where there is a question mark the author has made an educated guess considering the datasets previously used by those authors.

Frequency shows how often each data set mentioned a particular variable.

Significance

Chi squared tests were used on each variable to determine if it was significantly different from the average.

Manoeuvrability, Surprise, Command Control and Training were all above average and the Chi Squared test showed this was highly significant with a vanishingly small confidence level <0.0001%

Strong Arty/Air, Reserves, Logistics, Mission Urgency and Poor Weather were all below average effectiveness and the Chi Squared showing a <2% confidence level.

All other variables had a reasonable chance of being average with confidence levels >10%

We are comparing the variable weighting to the average effect, but the variables analysed had all ready been selected because they were more significant that others. In effect we compared to the average of the best variables rather than the average of all variables

This means that Manoeuvrability, Surprise, Command Control and Training have very significant effects on battle outcomes. Strong Arty/Air, Mission Urgency, Poor Weather, Reserves  have much less effect and are marginal, perhaps being no more effective than other unmentioned variables.

The other variables are effective but not as effective as the top four.

We can rank the variables on the basis of their chi squared scores with three being most effective and one the least

*There is a possibility of an overlap between some factors and own casualties. For example strong opposition may cause losses. Chi squared test between strong opposition and losses shows a confidence level of 0.02% and between Strong Arty/Air and casualties shows a confidence level of 1.5%. With the caveat that these comparison data sets are very small, this shows that these variables not associated, so the effect is likely psychological rather than physical. This certainly jibes with the anecdotal evidence on artillery and air which assign considerable psychological effects to these attacks. Also much firing is suppressive rather than destructive so again has a psychological dimension.

This seems to proves that these variables should be listed separately on the table, but a word of caution. There is only an overlap of three cases where the variables can be compared to losses, thus the sample size is rather small.

Frequency

As a further measure of legitimacy we can look at the frequency a particular variable is mentioned by the eight studies. The thesis being that the more important a variable the more often it will be mentioned.

Those that have a dark green cell in the frequency column were mentioned in 7 out of 8 studies and we can have a high degree of confidence that the historical record and professional opinion is in accordance as to the relative merits of these variables. It is difficult to put a threshold on where we should be concerned about lack of frequency but the reader should be concerned about whether bayonets and invulnerable armour are actually significant variables as they are only mentioned in two out of eight studies. The author believes they are significant when they happen but that they are rare occurrences. The dataset is US biased and the US was supplied with copious effective AT weapons hence invulnerable tanks is not often coded. However the effect is real as the battle of Rasiainai demonstrates, see Panzers of the Eastern Front edited by Peter Tsouras or Initial Period of the War on the Eastern Front by David Glantz. All the research on bayonet attacks indicates they are decisive if the attacker can close with the enemy (see below). Idealism is only mentioned in one study and its not surprising that the interviewees placed so much emphasis on this factor given they were ideological volunteers in someone else's war. This seems to be saying that idealism is a strong motivational factor for some soldiers but perhaps not all. https://www.nature.com/articles/s41562-017-0193-3 indicates that such factors figure prominently in both ISIS and PKK soldiers.

Discussion

It is pleasing to have a ranking of effects and the author is reasonably confident that the rankings are accurate given the data available.

There are some caveats. Firstly many of the datasets are questionnaire based, which means not all of the respondents got asked about every possible variable which could sway the results.

Apart from the Radey dataset the results are very US-centric. The Radey data set hints that these results may not be applicable in all combat situations.

Although  categorisation of most variables was straight forward, some were more challenging. For example Rowland's "aggressive recce" was coded under manoeuvre rather than surprise as it seemed from his description that seizing terrain was a more important function than information gathering.

The author tried to separate the effectiveness of a variable from the frequency with which it occurred. This is required if we want to get a set of wargames' modifiers, as the modifier should only applied after the pre-requisites are satisfied in the game. However it is far from clear from the datasets whether frequency enters into the judgements or not, especially those that are questionnaire based. It is fairly obvious the author has had limited success in this area. Bayonets are universally described as decisive if the attacker can close with the enemy. This implies that should be ranked as three rather than two. The rarity of bayonet charges accounts for its poor showing in the statistics.

One of the original aims of the research was to get a set of factors that could be used for morale tests in games. As the four top variables have little or no physical effect, the author believes it is safe to say they are mostly morale effects. 

A Note on Casualties

Most (all?) wargames rules base morale effects to a significant degree on casualties. However as we can see above there are at a host of other factors that produce greater morale effects than casualties.

To quote McQuie "To summarise, there seems to be no pattern of of influence. No matter how casualties are measured battles have been given up as lost when casualties ranged from insignificant to overwhelming"

Although many commentators have opined that between 25-33% casualties are enough to render a unit combat ineffective the evidence seems to indicate this is simplistic. This may be an average figure, but the variation around the mean is huge. Leonard Wainstein in http://professionalwargaming.co.uk/TheRelationshipBetweenBattleDamageAndCombatPerformance.pdf indicates for the battles that failed to produce victory, "Losses incurred from these formations covered a spectrum from virtually none to 56%, the heaviest being suffered by a unit moving ahead when relieved." In cases where a unit was victorious, "...formations ... continued to attack or defend despite losses of some degree from heavy to appalling"

The above examples are mostly at division level. However Dorothy Clark's report (about battalions) says, "The very wide individual differences in the ability of infantry battalions to carry out a given mission can not be accounted for in terms of casualties alone, no matter how the data are (sic) presented"

A Note on Bayonets

Very few bayonet casualties occur yet it appears the bayonet is a decisive weapon. Rowland in "Stress of Battle" reports that if a force can actually close with and over-run the enemy the enemy will surrender. However he is somewhat coy in explaining where this analysis came from.

Before WW1 it was a widely held belief, by most nations, that the bayonet attack was decisive. http://www.dtic.mil/dtic/tr/fulltext/u2/a437022.pdf discusses this. This quote is illustrative, "Victory is won actually by the bayonet, or by the fear of it, which amounts to the same thing so far as the conduct of the attack is concerned"

Dr John Stone has a very interesting analysis of bayonet fighting: https://kclpure.kcl.ac.uk/portal/files/80560627/A_Proxemic_Account_of_Bayonet_STONE_Firstonline22January2016_GREEN_AAM.pdf. His thesis, supported by numerous examples, is that bayonet fighting is very rare as either the attacking side halts before contact or the defending side breaks and runs. Only when there is mutual surprise between adversaries (i.e. in close terrain) does hand to hand combat take place. So in most cases, if the attackers are resolute the outcome is positive for the aggressor.

The number of bayonet charges has significantly diminished since WW2. Increased firepower and the tendency to react by dispersing has made closing with the enemy a very dicey proposition. http://www.dtic.mil/dtic/tr/fulltext/u2/692677.pdf shows the precipitous decline and http://www.thefirearmblog.com/blog/2016/09/08/age-bayonet-case-study-among-11-marines/ show there has been no upsurge in recent times.

Accounts of grenade usage in Korea (https://archive.org/details/DTIC_AD0000342) do not show the "make or break" effect of the bayonet charge. It seems that closing to zero meters, or at least the perception by the defending side that the attacker is prepared to do so, is critical in the outcome of the assault

This begs the question whether a charge without the bayonet, that closes to zero meters is still as effective as the bayonet charge? Unfortunately the author knows of no specific research around this matter but http://www.dtic.mil/dtic/tr/fulltext/u2/a612103.pdf hints strongly that the bayonet is required to produce the psychological effect on the enemy. Only one recorded encounter involved the bayonet, whereas there were multiple other close combat encounters with other weapons which implies the enemy were not scared and stood their ground. Even looking at encounters where only weapons were used (mostly rifles) the bayonet was used in 1 in 30 combats (3%).

In conclusion bayonet charges that are perceived by the defender as going to contact are decisive, though defensive fire must be overcome. It seem likely that the six inches of steel on the end of the infantryman's rifle are required to produce a decisive psychological impact.

What does this mean for Morale?

In the table below I attempt to distil morale factors suitable for wargaming out of the variables.

The first issue to address is that factors are likely to be different for skirmish games where one figure or counter equals one man and games where a counter or base represents multiple men. For example the loss of leaders will need to be tracked very carefully in a skirmish game, however in a larger scale game where a base represents a unit, the loss of leaders might be presumed to be related to the losses of men in some way. Thus the morale effect might become a function of losses.

Modelling Factors in a Wargame

In the work above we have compared factors and determined their relative effect in eroding or supporting morale. What we have not done is found a number for a break point. There are so many factors involved that essentially morale failure is a random process. We can say that some factors make morale failure more likely and we know that some factors are more powerful than others but in no way do we have the data to say unit Y under situation X will break Z% of the time.

There are two exceptions to this. British research on bombardments (see WO291/496) put a number on the weight of shells required to produce morale collapse in infantry, though the findings are somewhat tentative. Rowland states that if a unit successfully attacks another with bayonets it will almost certainly break. Unfortunately there is no way to extrapolate the likelihood of other variables producing (or preventing) morale failure from these two examples.

The percentages in the tables above suggest that manoeuvrability (i.e. attacks on on the flanks and rear) is one and half times more powerful in producing morale effects than the next variable. The other three third rank variables are about twice as powerful as the second rank variables. The relationship between the second and first rank variables is unclear. The percentage assigned to the idealism variable suggests that in some circumstances it can be significant.

There is some evidence that bayonets are under-rated in the table and the author is inclined to believe this should be a third rank variable.

When producing modifiers for a wargames' morale system the biggest dice roll modifier should be for flank and rear attacks. The next largest for surprise, training and leadership (command) and probably bayonet attacks. Other second rank modifiers should have smaller modifiers. The rules author needs to determine whether modifiers ought to be positive or negative. For example being in a leader's command radius could give a bonus. Alternatively losing a leader might give  penalty.

There are quite a large number of variables to encompass in a morale system so some could be "subcontracted" to other mechanisms. For example as described above, in a system where a base is a unit, leadership penalties could be subsumed into the combat results. The presumed loss of leaders in the base effecting how many casualties are required for its removal from the table as an effective fighting force. The morale effects of training and idealism might be subsumed into a "morale factor". Bayonets could be subsumed into the close combat mechanism.

Further Work

Although we have the relative merits of defeat mechanisms research on the absolute effects would be very useful to the rules writer, though this might be a pipe dream.

Work on the scale of variable required to produce a failure of morale. Losses effect morale but how many do you need for there to be a detrimental effect? How far away does an enemy unit have to be before it doesn't feel threatening etc.

Once morale is "broken" what are the likely outcomes?

Look at how this fits onto favourite rules mechanisms, in my case BBWW2B

The author right now does not have an "in" to the first three problems but there may well be a essay on the final one.

Conclusions

Defeat mechanisms are psychological or morale based. Although casualties are reasonably important, success depends on removing the enemy's resolve to keep fighting.

The four most important factors in defeat mechanisms are all purely morale based: command and control (leadership), surprise, manoeuvrability (getting on to the enemy flanks) and troop training. These factors have roughly double the effect of other modifiers this should be the primary ones modelled in a wargames' morale mechanisms. The next rank of defeat mechanisms are: strong opposition, morale, offensive firepower, losses, armour, exhaustion, invulnerable armour, bayonets and idealism.