Category Research & Analysis

The Key to Victory: Machine Learning the Lessons of History

Robert L. Helmbold has published a new book (his first book) called The Key to Victory: Machine Learning the Lessons of History. Bob Helmbold was one of the senior analysts at CAA (Center of Army Analysis). It was published by MORS (Military Operations Research Society) with the help of Dean Hartley, formerly of Oakridge. This is Bob Helmbold’s first book, and at 91 years old, I hope to see a dozen more from him.

Bob Helmbold will be doing a virtual presentation on this on the second day (Wednesday the 28th of September) of the Historical Analysis Annual Conference (HAAC).

The book is only available through Barnes and Noble, not through Amazon. A link to it is here: The Key to Victory: Machine Learning the Lessons of History: by Robert Helmbold, Paperback | Barnes & Noble® (barnesandnoble.com)

 

The First Historical Analysis Annual Conference (HAAC), 27-29 September 2022 in Tysons Corner, VA

Announcing the first Historical Analysis Annual Conference (HAAC), 27-29 September, 2022 in Tysons Corner, VA, USA. It is being hosted by The Dupuy Institute but is an open conference and we are looking for interested presenters and attendees from all corners of the historical analysis, operations research, and historical research communities.

Purpose:

To explore and promote the use of historical analysis in understanding military affairs.

It is built in part upon the work done by the HERO (Historical Evaluation Research Organization), TNDA (Trevor N. Dupuy and Associates), DMSI (Data Memory Systems Inc.), TMCI (The Military Conflict Institute), TDI (The Dupuy Institute) and elements of the Cornwallis Group. Similar in concept to the TMCI conferences.

First Conference:

Will be held 27-29 September 2022 at Tysons Corner, Virginia, near Washington, DC. It is in honor of the 60th anniversary of the Trevor N. Dupuy founded organizations and the 30th anniversary of The Dupuy Institute.

It is expected that the conference will include a wide range of attendees and presenters from private industry, academia, the U.S. government, the UK operational research and historical analysis community, and other interested parties.

The first day (Tuesday) of the conference with be focused on Analysis of Conventional Warfare, the second day (Wednesday) on Analysis of Unconventional Warfare and the third day (Thursday) with include Other Analysis of Warfare. Each presentation will be a maximum of 45 minutes with at least 15 minutes set aside for questions and discussion.

If demand is high enough, the conference will be broken into working groups as required.

Historical Analysis:

There is no clear definition as to what historical analysis consists of. For the purposes of this conference, it is that analysis of history that is focused on military affairs and is usually quantitative in approach and based upon a large number of cases (as opposed to being a case study). Historical Analysis is considered a proper subdiscipline of UK Operational Research. It is not considered part of Operations Research in the United States.

The UK Dstl defines historical analysis as “The use of mathematical, statistical, qualitative and other forms of analysis to understand historical engagements, operations, campaigns and conflicts for the purpose of providing impartial analysis and sensitive decision support to policy makers.”

The key elements of what The Dupuy Institute looks for in historical analysis is that it is 1) based upon history, with a strong bias towards primary sources (i.e. unit records), 2) it is a based upon a representative sample of cases, not just a case study of one or two cases, 3) it is analytical in approach, 4) it is using past real-world experience for analyzing and addressing a problem of today. Historical analysis is simply the analysis of real-world experiences.

There are people who are doing surveys of historical operations, basically doing multiple case studies to examine trends and patterns. While these are not based upon the large databases that The Dupuy Institute favors, these are efforts worth examining and such efforts will also be explored at our conferences.

Costs:

Cost of attendance will be $150 for the entire conference. Dining and hotels are at the expense of the attendees. There will be facilities for virtual attendees and virtual presenters, but the focus of the conference will be in-person presentations and attendees.

Call for Presentations:

We are making a call for papers and presentations at that time. We have already set up a preliminary list of presenters.

Long-term Goal:

The long-term goal is a create a series of periodic conferences for the purposes of presenting, examining and encouraging historical analysis of military affairs across a wide-range of issues.

For questions, suggestions, comments, or to volunteer for presentations, please contact:

Christopher A. Lawrence

The Dupuy Institute

(703) 289-0007

LawrenceTDI@aol.com

 

P.S. In subsequent posts I will be posting the rates for the conference, the list of nearby hotels, and a preliminary list of presenters.

JSTOR, Trevor Dupuy, Combat Data and the 3:1 Rule

In moments of quiet I sometimes search the internet to see if people are referencing our work. Sometimes I run across articles and discussions I have forgotten about. This was one of them: Combat Data and the 3:1 Rule

We have blogged about this subject a few times before (and even referenced the JSTOR article):

The Source of the U.S. Army Three-to-One Rule | Mystics & Statistics (dupuyinstitute.org)

The U.S. Army Three-to-One Rule versus the 752 Case Division-level Data Base 1904-1991 | Mystics & Statistics (dupuyinstitute.org)

The U.S. Army Three-to-One Rule versus 49 U.S. Civil War battles | Mystics & Statistics (dupuyinstitute.org)

The U.S. Army Three-to-One Rule versus 243 Battles 1600-1900 | Mystics & Statistics (dupuyinstitute.org)

The U.S. Army Three-to-One Rule | Mystics & Statistics (dupuyinstitute.org)

The Great 3-1 Rule Debate | Mystics & Statistics (dupuyinstitute.org)

The 3-to-1 Rule in Recent History Books | Mystics & Statistics (dupuyinstitute.org)

Questioning The Validity Of The 3-1 Rule Of Combat | Mystics & Statistics (dupuyinstitute.org)

Comparing the RAND Version of the 3:1 Rule to Real-World Data | Mystics & Statistics (dupuyinstitute.org)

TDI Friday Read: The Validity Of The 3-1 Rule Of Combat | Mystics & Statistics (dupuyinstitute.org)

The 3-to-1 Rule in Histories | Mystics & Statistics (dupuyinstitute.org)

Trevor Dupuy and the 3-1 Rule | Mystics & Statistics (dupuyinstitute.org)

Accounting for Death in War

Just spotted on Twitter that Dr. Michael Spagat is offering a free course Accounting for Death in Warfare: Separating Fact from Fiction. It starts 24 May. I gather it is free and anyone can sign up. See: www.futurelearn.com

I am a big fan of Michael Spagat’s work, having blogged about it before:

TDI Friday Read: Mike Spagat’s Economics of Warfare Lectures & Commentaries | Mystics & Statistics (dupuyinstitute.org)

Anyhow, I did confirm with Dr. Spagat that it is open to anyone and free. 

The ORO/RAC/GRC Library

I noticed my posts in 2017 on the HERO and SAIC Libraries got a couple of hits a week or so ago. I was actually planning on continuing posting on this subject at the time, but got sidetracked. So, let me take a moment now to post on another disappeared library.

The Operations Research Office (ORO) was the Army’s RAND. Formed in 1948 under the leadership of Ellis Johnson, it grew to considerable size (220 in June 1952) providing studies and analysis for the U.S. Army. ORO  was tied to Johns Hopkins University. But, the leadership of ORO had a contentious relationship with the army, so the Army decided in 1961 to dissolve it and reconstitute it under new leadership. The army set up a new corporation called Research Analysis Corporation (RAC) and hired back everyone at ORO but the senior management. It was also claimed that RAC stood for “Relax and Cooperate.”

RAC continued the work of ORO and in fact they were functionally the same organization. Under these two organizations, it included developing the first Army combat models (Carmonette and Atlas) in addition to field analysis of the Korean and Vietnam wars. They also had a historical analysis capability, having employed Dr. Hugh Cole,. the person who managed and wrote part of the Army “Green Book” series, the official history of the U.S. Army in World War II.

The Army decided to then build its own RAND. This was the Concepts Analysis Agency (CAA) founded in January 1973. My father was one of army officers assigned to this new outfit, which is the connection that eventually led me into this business. As I had a strong interest in military history and wargaming as a teenager, my father working at the Army’s premier wargaming shop got my interest. I often discussed with him the work they were doing, and in particular, his rather negative evaluation of the combat models they were using (in particular CEM – Concepts Evaluation Model). The Ardennes Campaign Simulation Data Base (ACSDB) that I managed in 1987-1989 was for the purpose of validating CEM. 

With the Army having created its own RAND, RAC as an independent FFRDC (Federally Funded Research and Development Center) was no longer needed. They therefore mutated in a private contractor called General Research Corporation (GRC). They inherited most of the staff and the ORO/RAC library. Dr. Hugh Cole retired but ended up working for us on the ACSDB as a consultant, along with Charles McDonald.

GRC continued doing business, although I do not know what that was, as I never intersected with them. I started working in this business in 1987 at Trevor Dupuy’s DMSI (Data Memory Systems, Inc.) doing contract work for CAA among others. CAA eventually changed its name to Center for Army Analysis (CAA).

GRC about 15 or so years ago (operating from memory here) was bought out by AT&T defense systems (AT&T = old Bell Telephone). I also had never intersected with them. They inherited the ORO/RAC/GRC library and took over the GRC building in Tysons Corner. When I heard that they had shut the library down, I payed a visit to their office.

It turns out that this library, built up over 50+ years of research, was indeed shut down. It was taking up space and they decided they needed another conference room! So, they told the employees they could take whatever they wanted from the library and threw the rest away. So, the library and all of its material was dissolved and thrown away.

So… where does one find ORO/RAC/GRC material? Well, I have a rather thick bibliography listing all the ORO/RAC reports. Forget how I scored that. It is a unique item and I don’t know of anyone else who has one. On the other hand, if you ask me to find it, I don’t know in which of the 80+ boxes of books it is in that I have stashed away. I am not going to even try at this point to find, as they are stacked up (although it is good exercise). Of course, ORO/RAC/GRC submitted their reports to DTIC and NTIS. But, as I pointed out in the my post on TDI Reports at DTIC that only about 40 percent of HERO reports are listed in DTC and 0 of the 80 TDI reports are listed there. Now, I also have all the HERO reports in those stacks of boxes I have stashed away. A friend of mine at CMH (Center for Military History) did inform me that some/all of the ORO reports are sitting in archived files of one DOD agency in the national archives, but these section of files are still classified and not available to most researchers. Also, they are difficult to locate and access. CAA also has its own library, but I do not know the state or extent of it. If CAA needs a HERO/DMSI/TNDA/TDI report, they invariably end up asking me, which does not give me a lot of confidence. I did ask a senior analysis at CAA once where he would go to get an ORO/RAC report, and he told me he would just call the library at GRC. Obviously that is no longer an option. 

I do not know the state of the CAA library or its holdings. When the GRC library was shut down I did flag the issue to the director of CAA. The corrective action I believe would have been to assemble a small team to create a definitive Army Studies and Analysis Research Library to collect all the ORO/RAC/GRC and HERO/DMSI/TNDA/TDI and CORG/CDEC/CAA and RAND Arroyo and other relative studies in one place (probably the CAA library). This effort was not acted on. 

It is probably still possible to do at this point, but as the “graybeards” are disappearing from the business, and it is amazing what they keep stored in their files cabinets and basements, then it will get harder to do over time.

 

Related Posts:

TDI Reports at DTIC | Mystics & Statistics (dupuyinstitute.org)

The HERO Library | Mystics & Statistics (dupuyinstitute.org)

The SAIC Library | Mystics & Statistics (dupuyinstitute.org)

Less related posts:

Company Commander | Mystics & Statistics (dupuyinstitute.org)

Validation Data Bases Available (Ardennes) | Mystics & Statistics (dupuyinstitute.org)

Historians and the Early Era of U.S. Army Operations Research | Mystics & Statistics (dupuyinstitute.org)

The Use of the Two Campaign Data Bases | Mystics & Statistics (dupuyinstitute.org)

Some sources:

A History of the Department of Defense Federally Funded Research and Development Centers (I am actually the author of this, although I am listed as a researcher): https://www.princeton.edu/~ota/disk1/1995/9501/9501.PDF and here: https://digital.library.unt.edu/ark:/67531/metadc39765/m1/1/

History of Operations Research in the United States Army, Volume I: 1942-1962: CMH Publications Catalog – HISTORY OF OPERATIONS RESEARCH IN THE UNITED STATES ARMY, VOLUME 1: 1942-1962

 

 

 

Wargaming in the U.S. Army

I was asked recently about wargaming in the U.S. Army, and I kind of punted on the question. I then got a friend of mine to answer it. The wargaming efforts we have been involved in have been primarily for analysis and casualty estimation efforts. We have not been involved in wargaming for training, officer development, or development of planning. Other people have been doing that, I gather with mixed results. Here is two recent articles on the subject that were forwarded to me:

1. Educational and Tactical Use of Wargames with the U.S. Army: https://warroom.armywarcollege.edu/wargaming-room/tactical-edge/

2. Wargaming the Gray Zone:

https://warontherocks.com/2021/03/whispers-from-wargames-about-the-gray-zone/

 

Some Background on TDI Data Bases

The Dupuy Institute (TDI) are sitting on a number of large combat databases that are unique to us and are company proprietary. For obvious reasons they will stay that way for the foreseeable future.

The original database of battles came to be called the Land Warfare Data Base (LWDB). It was also called the CHASE database by CAA. It consisted of 601 or 605 engagements from 1600-1973. It covered a lot of periods and lot of different engagement sizes, ranging from very large battles of hundreds of thousand a side to small company-sized actions. The length of battles range from a day to several months (some of the World War I battles like the Somme).

From that database, which is publicly available, we created a whole series of databases totaling some 1200 engagements. There are discussed in some depth in past posts.

Our largest and most developed data is our division-level database covering combat from 1904-1991 of 752 cases: It is discussed here: The Division Level Engagement Data Base (DLEDB) | Mystics & Statistics (dupuyinstitute.org)

There are a number of other databases we have. They are discussed here: Other TDI Data Bases | Mystics & Statistics (dupuyinstitute.org)

The cost of independently developing such a database is given here: Cost of Creating a Data Base | Mystics & Statistics (dupuyinstitute.org)

Part of the reason for this post is that I am in a discussion with someone who is doing analysis based upon the much older 601 case database. Considering the degree of expansion and improvement, including corrections to some of the engagements, this does not seem a good use of their time., especially as we have so greatly expanded the number engagements from 1943 and on.

Now, I did use some of these databases for my book War by Numbers. I am also using them for my follow-up book, currently titled More War by Numbers. So the analysis I have done based upon them is available. I have also posted parts of the 192 Kursk engagements in my first Kursk book and 76 of them in my Prokhorovka book. None of these engagements were in the original LWDB. 

If people want to use the TDI databases for their own independent analysis, they will need to find the proper funding so as to purchase or get access to these databases. 

“Our enemies generally shoot first 80 percent of the time”

Someone just asked me about a quote in the following article:
https://breakingdefense.com/2020/11/army-wants-smaller-brigades-stronger-divisions-lots-of-robots/

The quote is:

“Historical data on direct-fire engagements “shows that our enemies generally shoot first 80 percent of the time,” Sando said. “We don’t like those odds, [so] we want to avoid the close fight if we can. If we can’t avoid it, we want to enter it under conditions that are favorable to us.”

The question was: did this statistic come from us or do we know where it came from? Well, it definitely did not come from our work or Trevor Dupuy’s work. I have never heard of such a claim before.

Does anyone know where this came from? Is there some study or piece of quantitative historical analysis that this claim is based upon. I have my doubts about the accuracy of this claim.

If the data was drawn from unconventional (recent) warfare scenarios, then by nature of those conflicts, the insurgents initiate engagements more often than the counterinsurgents do. Not sure it would be as high as 4-to-1.

If it was drawn from conventional war scenarios, then by their nature it should be 50/50, unless you are always the attacker. Even then I am not it would be as high as 4-to-1.

 

TDI and the TNDM

The Dupuy Institute does occasionally make use of a combat model developed by Trevor Dupuy called the Tactical Numerical Deterministic Model (TNDM). That model is a development of his older model the Quantified Judgment Model (QJM). 
 
There is an impression, because the QJM is widely known, that the TNDM is heavily involved in our work. In fact, over 90% of our work has not involved the TNDM. Here a list of major projects/publications that we done since 1993.
 
Based upon TNDM:
Artillery Suppression Study – study never completed (1993-1995)
Air Model Historical Data feasibility study (1995)
Support contract for South African TNDM (1996?)
International TNDM Newsletter (1996-1998, 2009-2010)
TNDM sale to Finland (2002?)
FCS Study – 2 studies (2006)
TNDM sale to Singapore (2009)
Small-Unit Engagement Database (2011)
 
Addressed the TNDM:
Bosnia Casualty Estimate (1995) – used the TNDM to evaluate one possible scenario
Casualty Estimation Methodologies Study (2005) – was two of the six methodologies tested
Data for Wargames training course (2016)
War by Numbers (2017) – addressed in two chapters out of 20
 
Did not use the TNDM: 
Kursk Data Base (1993-1996)
Landmine Study for JCS (1996)
Combat Mortality Study (1998)
Record Keeping Survey (1998-2000)
Capture Rate Studies – 3 studies (1998-2001)
Other Landmine Studies – 6 studies (2000-2001)
Lighter Weight Armor Study (2001)
Urban Warfare – 3 studies (2002-2004)
Base Realignment studies for PA – 3 studies (2003-2005)
Chinese Doctrine Study (2003)
Situational Awareness Study (2004)
Iraq Casualty Estimate (2004-2005)
The use of chemical warfare in WWI – feasibility study (2005?)
Battle of Britain Data Base (2005)
1969 Sino-Soviet Conflict (2006)
MISS – Modern Insurgency Spread Sheets (2006-2009)
Insurgency Studies – 11 studies/reports (2007-2009)
America’s Modern Wars (2015)
Kursk: The Battle of Prokhorovka (2015)
The Battle of Prokhorovka (2019)
Aces at Kursk (2021)
More War by Numbers (2022?)
 
 
Our bread and butter was all the studies that “did not use the TNDM.” Basically the capture rate studies, the urban warfare studies and the insurgencies studies kept us steadily funded for year after year. We would have not been able to maintain TDI on the TNDM. We had one contract in excess of $100K in 1994-95 (the Artillery Suppression study) and our next TNDM related contract that was over $100K was in 2005.
 
  

Amphibious and River Crossing Engagements in the Italian Campaign 1943-44

Now, there were a few amphibious operations and major river crossing operations during the Italian Campaign. These are significant enough that we should examine them separately. Among the 141 Italian Campaign engagements there are three amphibious operations and 16 river crossing operations. These are outlined in our previous post:

Amphibious and River Crossing attacks in Italy 1943-44

This is a small number of cases, but does seem to show that the defender has an advantage when defending against amphibious assaults and river crossings. To compare them to the rest of the engagements:

 

Amphibious Operations Comparison – Italian Campaign 1943-44:

………………………………………………………………..Average..Average

……………………………………………………Percent…Percent…Attacker..Defender

………………………………….……..Cases….Wins…….Advance..Losses…Losses

U.S./UK Attacking – Amphibious…..3……….100………..100………..417……….105

UK Attacking – other RM…………….10………..40………….67…………382………152

U.S. Attacking – other RM…………..20………..65………….80…………341………430

 

………………………………………….……………Force….Exchange

………………………………….……..Cases…….Ratio…..Ratio

U.S./UK Attacking – Amphibious…..3……………3.45…….3.98-to-1

UK Attacking – other RM…………….10…………..1.94…….2.51-to-1

U.S. Attacking – other RM…………..20…………..1.69…….0.79-to-1

 

River Crossings Comparison – Italian Campaign 1943-44:

………………………………………………………………..Average..Average

……………………………………………………Percent…Percent…Attacker..Defender

………………………………….……..Cases….Wins…….Advance..Losses…Losses

UK Attacking – River Crossing……8………….63………….67………..169………154

UK Attacking – other FM…………….6………….50………….67………..306………284

U.S. Attacking – River Crossing…..6…………67………….67…………388……..112

U.S. Attacking – Other RM………..17…………59………….76…………374……..477

German Attacking – River Crossing..2……….50………..100…………955………507

German Attacking – Other FM………5………..40………….40…………756………805

 

………………………………………….……………Force….Exchange

………………………………….……..Cases…….Ratio…..Ratio

UK Attacking – River Crossing……8……………2.20…….1.10-to-1

UK Attacking – other FM…………….6……………1.71…….1.08-to-1

U.S. Attacking – River Crossing…..6…………..1.47…….3.47-to-1

U.S. Attacking – Other RM………..17…………..1.72…….0.78-to-1

German Attacking – River Crossing..2…………1.96…….1.88-to-1

German Attacking – Other FM………5…………1.39……..0.94-to-1

 

Now this is a very small number of cases for comparison, and just for balance I tried to match it across similar terrain types.[1] Still, even with this very limited data, there appears to be a clear pattern. In every case the exchange ratio more favors the defender. In the case of amphibious operations, this either 2.51-to-1 or 0.79-to-1 compared to 3.98-to-1 for amphibious operations. For river crossings this is 1.08-to-1 compared to 1.10-to-1 for UK river crossings, 0.78-to-1 compared to 3.47-to-1 for U.S. river crossings (or 0.78-to-1 compared to 0.82-to-1 if the two Rapido River engagements are left out), and 0.94-to-1 compared to 1.88-to-1 for German river crossings. Obviously a whole lot more cases would need to be examined to establish anything with certainty, but there are not a lot of amphibious operations over river crossings coded in the ETO or Kursk/Kharkov sections of the database. It would require a significant research effort to assemble such data.

Of course, what this really all points to is that the data needs to be more tightly prescribed and a whole lot more cases need to be assembled.

 

 

———

[1] In the case of U.S./UK Attacking – Amphibious, one engagement was done by the U.S. and the other two by the UK. In the case of UK Attacking – River Crossing there were 7 cases in flat mixed terrain and 1 case of rolling mixed terrain. In the case of U.S. Attacking – River Crossing there were 3 cases of rolling mixed terrain, one case of Rugged Mixed, and two cases codes as FB/M, meaning the terrain varied from flat bare to flat mixed. In the case of German Attacking – River Crossing, both engagements were in flat mixed terrain but one was against the U.S. and the other was against the UK. In the case of German Attacking – Other FM two engagements were against the U.S. and three were against the UK.