I've read an article recently (https://arxiv.org/abs/1805.00463) that claims that computer games are often useful in teaching history courses, and a recent article from MIT Technology Review claims that they may be useful in research, sparing something they call 'computational history'.
How useful are simulations and computational tools in the study of history in general, and these games in particular?
"Computational history" doesn't have a universal precise definition, but it can be less precisely defined as the application of computational methods to history. It is dominated by methods for dealing with data:
Digitisation of texts (including OCR).
Machine translation of texts.
Searching of texts.
Machine analysis of the content of texts: text analysis, text mining, data mining, machine learning, deep learning, etc.
The first two are about making data accessible. The third is about helping historians make use of that data (if you have 10 million pages of documents, on which pages is the useful stuff?). The last is about using computational techniques for making sense of large amounts data (if you have 10 million pages of documents, what do they say?).
For an overview of this data-oriented part of computational history (i.e., most of computational history), see:
Bojan Bozic, et al. (eds), Computational History and Data-Driven Humanities, revised selected papers from Second IFIP WG 12.7 International Workshop, CHDDH 2016, Dublin, Ireland, May 25, 2016, Springer, 2016. https://doi.org/10.1007/978-3-319-46224-0
Shawn Graham, Ian Milligan and Scott Weingart, *Exploring Big Historical Data: The Historian's Macroscope", Imperial College Press, 2015. https://doi.org/10.1142/p981 http://www.themacroscope.org/2.0/
There is a lot of potential for simulations in computational history. What simulations need that isn't present just in data (even if there is a vast amount of data) are models. With a model, predictions can be made. Predictions can be used for:
The validation of data. If a prediction based on a source of data doesn't match reality, and the model is well-tested and validation for the range of that data, then that data might be wrong (or incomplete).
To extend limited data. E.g., if the full (unknown) data should be accurately described by a power law, the missing part of the full data can be estimated from the data that is available.
To explore counter-factual histories. For a down-to-Earth example, consider simulations of a battle. Such simulations can be used to determine the probability of different outcomes. If the historical outcome was overwhelmingly likely, then no special circumstances are needed to explain it - it follows from the already known data.
And, of course, predictions are essential for validating the models in the first place, before the above can be carried out. The last of the data-oriented techniques, machine learning, can be used to construct models. Sometimes (often!) models produced in this way are opaque - the model is not understood by the humans that the machine delivers the model to. However, a not-understood empirical model like this can still be used, as a mystery black box that can produce reliable predictions.
Compared to Big Data computational history, simulations are only a very small part of the field. They have the potential to provide a lot of useful results, even if they remain a very small part of the field. For more on simulations, see
Both data-oriented computation and simulations have a role to play in archaeology, which is often an important source of data for history (and vice versa!). The data-oriented applications are very similar: making data available digitally, making data searchable, and exploring data with machine learning techniques. Rather than scanning pages of text (a common method for digitising data in history), artifacts can be photographed, or laser scanned, or other analysis can be performed. The simplest version - already providing a valuable resource - is for museums to have online searchable collections, with descriptions and some data (preferably including photos).
Simulations in archaeology can generally be described as computational experimental archaeology. Experimental archaeology involves things like learning about the function of artifacts by testing them (or replicas) in action. Computational modelling and simulation can allow many variations of a particular artifact to be tested (in simulation), which provides information about how optimum the design is, and how effective the design is. For example, questions such as the penetration of armour by arrows can be studied in simulation: