Military Strategy Magazine  /  Volume 7, Issue 4  /  

Drones in the Nagorno-Karabakh War: Analyzing the Data

Drones in the Nagorno-Karabakh War: Analyzing the Data Drones in the Nagorno-Karabakh War: Analyzing the Data
Image from Radio Free Europe at - Still released by Azerbaijan's Defense Ministry, October 1, UAV.
To cite this article: Hecht, Eado, “Drones in the Nagorno-Karabakh War: Analyzing the Data,” Military Strategy Magazine, Volume 7, Issue 4, winter 2022, pages 31-37.

The use of remotely piloted or autonomous aircraft, from now on called ‘drones’, has increased dramatically over the past two decades and has generated a debate on whether they are merely one more tool of war or a revolution in warfare. This debate escalated during the Second Nagorno-Karabakh War, but most articles simply stated opinions without providing actual data to support them.

Many accounts of the war describe it as a one-sided, an Azeri drones versus Armenian ground forces event with Azeri ground forces figuratively “riding on the backs” of the drones to victory with a minimum of fighting by them. To quote a typical example:

“Azerbaijan’s UAVs obliterated Armenia’s formidable array of ground-based air defences, after which they systematically decimated Armenia’s ground force matériel, including tanks, artillery pieces, and supply trucks. This onslaught forced Armenia to accept a humiliating ceasefire imposed by Russia.

… It can be said that this was the first postmodern conflict, in that it was the first in which unmanned-aircraft overwhelmed a conventional ground force, grinding it down to the point of impotence and paving the way for the Azeri ground forces to roll in and take possession of a strategic chokepoint.”[i]

Do the available numbers support statements such as this?

And if so – were drones necessary to achieve this result, or could it have been achieved by ‘ordinary’ aircraft?

Nagorno-Karabakh in Numbers

First a qualification – neither side in the conflict has released reliable numbers. Reading the daily claims of both sides during the war clearly shows exaggeration and misdirection by both.[ii] Numbers published daily contradicted numbers published previously and later.

Since only Azerbaijan employed armed drones, both munition-dropping and suicide versions, the focus is on the capabilities and limitations exposed by them. However, beyond the technical aspects of the drones themselves, there are tactical and professional aspects on the Azeri side that may have prevented them from fully exploiting drone capabilities and technical, tactical, and professional aspects on the Armenian side[iii] that may have assisted the Azeris in achieving more than they would have against a better-prepared foe. A data study collecting all the publicly available video and photographs of destroyed Armenian equipment, separating proven drone-kills, proven kills by other-weapons and kills by unknown weapons, has been published by an independent research team named Oryx,[iv] but it contains fewer targets than the total claimed destroyed by the Azeris. Perhaps the proven destructions by drones are fewer than the total destroyed by drones, and perhaps the number of photographically proven drone-destructions is virtually all there were. The rest were destroyed by other weapons that had no photographic back-up. In any case, the following analysis must be treated with caution. The opposite is more accurate, in that all claims for or against the future of drones based on this war do not utilize the available data, so are less reliable.

After the war, President Aliyev published a summary of Armenian equipment destroyed and captured by the Azeris.[v] For our purposes only the destroyed equipment matters. Some of the captured items were damaged and some were abandoned undamaged, but there is no account separating the two. The following table compares Aliyev’s statement with the Oryx video and photograph collection.

It should be remembered that the photographic sample provided by the Azeris shows only the successes – never the misses.

The total photographic sample covers nearly 60% of Aliyev’s claim and 75% of the sample was destroyed by drones, i.e., almost 45% of the total claimed by Aliyev were definitely destroyed by drones. Even assuming that this is the complete portion of targets destroyed by drones, this is certainly a sizeable proportion. However, the actual proportion might be larger since we do not know how many more items were destroyed by drones without publication of photographic evidence. As far as the destroyed trucks and most other soft skinned-vehicles are concerned, it is likely to be almost all of them, given the locations they were destroyed – some distance from the front lines. It is also unlikely that they were targeted by artillery in those locations. The Azeri air force did conduct approximately 600 sorties by manned-aircraft,[vii] mostly Su-25s and attack-helicopters, but there is only anecdotal information on their targets.

Ostensibly the claims of Azeri ground forces riding to success on the back of a storm of drones are vindicated. However, this conclusion is complicated by other data.

First and foremost, the casualties suffered by the Azeris, which is a minimum of 2,900 admitted killed and a few thousand wounded.[viii] This was not a ground force that fought a battle made easy by the effects of massive drone strikes. This ground force had to fight casualty-intensive battles to defeat a determined enemy, no less well equipped and no less proficient than itself – a peer enemy. The drones definitely tipped the balance in favour of the Azeris, but by themselves, they did not win the war – not even close.

Furthermore, 563 certain destructions by drone’s average to only 13 per day of the 44 day war. Adding 75% of the equipment claimed destroyed by Aliyev (assuming the ratio of targets destroyed by drones to total targets destroyed is the same as the photographic sample) raises this to only 22 targets per day. Reducing days on which there were no drone strikes (at least four such days, according to the Armenians). There were 14 days in which the Azeris did not publish new drone-strike videos (though whether because they had none or chose not to is not known), and concentrating more strikes on particular days to fit the waxing and waning of the ground combat and the vagaries of official Azeri statements, does not suggest an overwhelming rate of destruction. Furthermore, declared Armenian fatalities are less than 1.5 times those of the Azeris (currently almost 4,000 dead and missing), but comparing equipment losses is almost impossible as there is very little photographic evidence. The Azeris have not provided any numbers on their equipment losses and Armenian claims seem grossly exaggerated (784 tanks and other AFVs by morning of 8th November[ix]).

The fact is that on the first days of the war repeated Azeri ground attacks failed to penetrate Armenian defences[x] and that even after they finally succeeded, exploiting this success faced stiff resistance and they suffered a few more tactical defeats before the final victory. The war was won by Azeri perseverance in the face of heavy casualties and many small defeats while gradually wearing-down Armenian forces no-less determined than the Azeris[xi] and gradually taking ground till the Armenian political and military leadership realized that the situation was irretrievably lost, and further resistance would cost more casualties and territory but achieve nothing.

Perusing the Azeri Ministry of Defence statements suggests that on the first days the drone force focused on destroying Armenian air defences. Based on their official statements, strikes on air defences continued throughout the war at a slower pace, suggesting that the Azeris were satisfied with the initial results. However, the strike videos they released showed much fewer air defence targets struck than declared – so either the declarations were exaggerated, or the videos were only a chosen sample. The numbers declared accumulated gradually to 61 air defence targets on 7th October (Day ten of the war), but then, on 9th October, they reduced the accumulated total to only 27 and gradually added more until Aliyev’s final statement of 73 items all together.

A similar pattern can be seen also in regard to the variety of other targets attacked: publication of statements much higher than the video evidence; an accumulation of enemy targets destroyed reaching a peak at noon 9th October, and that evening a reduction in the accumulated claims (tanks and other AFVs from 275 to 232 and artillery systems from 286 to 242) and a much more gradual accumulation from then to the end of the war. The number of strike videos released always trailing behind the textual claims. It should be remembered that not all these targets were hit by drones.

Perusing the photographic evidence of drone strikes suggests Azeri preferences in attacking targets: nearly twice as many artillery targets were struck than tanks and other AFVs. Trucks are about 28% of the photographic sample of drone-destroyed targets – more than tanks and AFVs (22%) but less than artillery (38%). Adding the other types of soft-vehicles struck and taking into account that even without photographic evidence, the location of most of the trucks and vehicles when destroyed or damaged was in areas that strongly suggest they too were hit by drones or other aircraft, changes the proportions but still does not necessarily change the order of priority the Azeris apparently ascribed to the different target types. What is clear is that the majority of the drone strikes were not against Armenian forces in the front line – the focus is on the artillery support, the armoured reserves and the logistics (not only trucks – also supply bases, especially ammunition storage).

To conclude: it is very clear that without the drones the Azeris would not have achieved the success that they did. However, it is just as clear that the drones did not win the war by themselves and did not make the ground battle easy. Given the available data, computing the exact share in victory between drones and ground forces more accurately than that is impossible.

Failure of Armenia’s Air Defence

By war’s end, the Armenians claimed to have shot down a grand total of 264 drones, 25 combat aircraft and 16 helicopters,[xii] however, provided no evidence. If these numbers are true, then the Armenian air defence is definitely worthy of the adjective “formidable” as quoted above. The Azeris deny anything close to these numbers but provide no real numbers or evidence of their own. What is clear is that even if the Armenians did shoot down 264 drones, the Azeris apparently had many more available – enough to achieve the results described above. There is no way to provide other numbers, but the achievements of the Armenian drones and air force in general suggest that the Armenian claims are a gross exaggeration.

To paraphrase British Prime Minister Stanley Baldwin argument in 1932, “The drone will always get through”?[xiii] Or, at least enough of them to ultimately make the defensive futile?

This should not be inferred from the Second Nagorno-Karabakh War. A careful study of Armenia’s air defence shows that it was not “formidable”, certainly not where drones are concerned. Exact amounts of the various missile systems are not available, but they included a combination of Strela-10 (SA-13), Osa (SA-8), Kub (SA-6), Krug (SA-4), S-300 and Tor. Except the Tor, all were older less capable versions.

Only the Tor was a threat to the Bayraktar TB-2 and Israeli-made suicide-drones.[xiv] The effective range of the Strella-10s and Osas against the TB-2 drone sized targets was shorter than the range of the missiles these carried; whereas the longer-ranged Kubs, Krugs and S-300s were optimized against targets bigger and faster than the drones, so to them the drones were invisible.[xv] Apparently there were only 6 Tors. Given the overall size of Nagorno-Karabakh and its mountainous terrain, 6 Tors were a drop in the bucket compared to the number needed to create a robust defence with overlapping fields-of-fire to provide cover for each other. One Tor was destroyed towards the end of the war. The Azeris observed it with a drone from a safe distance till it folded its antenna and drove into a garage for maintenance or rest. As soon as it was unable to defend itself, it was bombarded with a number of suicide drones. Not only were the Armenians lacking in numbers of relevant systems, but neither did they use those they had properly – sending them alone rather than providing each other cover. So basing computations of the future capability of drones against air defences on the Nagorno-Karabakh war is misleading. It cannot be assumed in advance that future enemies will be as weak as the Armenians were.

An important question is why did the Armenians not acquire better systems – this war was not the first time they had faced drone strikes launched by the Azeris. The Azeris had used Israel-made suicide drones in a number of previous skirmishes since 2016. However, apparently the Armenians believed they were protected. After a four-day skirmish in July 2020, an Armenian Major-General stated that during that skirmish: “… the Armenian army destroyed more than a dozen Israeli strike drones that were in the Azerbaijani arsenal within a matter of days. These drones were made of the best technology and they were considered indestructible.”[xvi] The only change made by the Azeris from the July skirmish to the war, was to add a new drone to their arsenal – the missile-firing Bayraktar TB2 drone, enabling them to strike targets up to 8 kilometers away. The vast majority of the strike videos released by the Azeri Ministry of Defence were filmed by the TB2s, but these include videos of suicide-drone strikes, so it is not quite clear how many of these videos show actual TB2 strikes or the TB2 is merely the spotter for the suicide-drone attack. If the new missile-firing drone is the reason for the change in level of success, it would suggest that the Armenian air defences had been perhaps sufficiently effective in shooting-down of suicide-drones, which need to approach the target and therefore operate deep in the defensive envelope of the defensive systems, including ordinary anti-aircraft guns. An alternative explanation is that the Azeris had used their suicide-drones sparingly, so the Armenians were lulled into complacency by their presumed success in defeating this weapon.

One weakness of the remotely piloted drone is the threat that the enemy might override the controls and force it to crash by jamming or spoofing the signals sent by its pilot. One report claimed that 9 Azeri drones had been brought down in this manner when they flew too near a Russian army base in Armenia.[xvii] After the war, an Armenian general stated that the Armenians had successfully used a Russian electronic warfare system for several days. He did not specify whether the system brought down the drones or just forced them to maintain pilot control.[xviii]

The lesson is clear – armies must develop and procure large numbers of anti-drone capable systems. Systems optimized to confront manned-aircraft are usually not sufficient to confront the smaller drones – though against larger drones they can be effective.[xix]

Some analysts have suggested the issue was the lack of Armenian combat-field-craft – they were parked or drove in the open with insufficient use of camouflage or terrain concealment and often were too tightly bunched into a convenient target to be detected and attacked. However, better combat-field-craft would not have solved the Armenians’ problem. The terrain over most of the theater is devoid of tall vegetation or other options of concealment. Furthermore, some strike videos clearly show failed attempts to conceal equipment in small woods or under camouflage nets. The ability of the drones to conduct long sweeps of an area with multi-spectral cameras enabled them to find these targets too. Furthermore, concealment prevents movement – how would the Armenians have brought up reinforcements or conducted counter-attacks while hiding? Also, the concept of spreading out against small guided steep trajectory munitions is almost irrelevant – unlike statistical or flat-trajectory munitions, the miss, if it occurs will be very close to the target, and the warheads are fairly small.

The only solution is to provide active interception of the drones and the munitions – an ‘interception dome’ of mobile weapons that can cover an area large enough for a ground forces company or battalion to maneuver in and can move with that unit to maintain that dome wherever it goes. Electronic warfare is useful but might accidentally bring down friendly drones, whereas interception weapons can be equipped to discern friendly from hostile drones. Another issue is that whether using physical interceptors or electronic warfare, the defending unit continuously signals its own location and that of the unit it is defending to the enemy’s signals intelligence.


What can be learned from this war on the topic of drones?

First a qualification – an issue not discussed here is the tactical effect of drone-swarms as opposed to single drones, since none were employed.

Second – it is clear that the hype was exaggerated. The Azeri drones were essential for their victory, but did not win the war alone, severe ground fighting was necessary.

Some of the lessons are not new – when one side has an advantage in the air, he gains a considerable advantage on the ground too. To quote Erwin Rommel, who faced manned-aircraft – not drones: “Anyone who has to fight, even with the most modern weapons, against an enemy in complete command of the air, fights like a savage against modern European troops, under the same handicaps and with same chances of success.”[xx]

The Azeris did not command the air, but the ability of their drones to exploit a specific gap in the Armenian air defence, gave them freedom to use the air and gradually, as they destroyed more and more Armenian air defence assets, provided operational freedom to use manned-aircraft too. However, this gap was created by Armenian mistakes, not by the essential nature of drone warfare. Furthermore, though the gap can probably only be reduced, not fully closed as with the use of even smaller drones by the Islamic State and other organizations. The converse is that the smaller the drones needed to exploit what remains of that gap, the smaller the size of the munitions they can carry and therefore the smaller their tactical effect. In fact, most of the drones today can carry munitions equivalent only to attack-helicopters. Whenever a bigger bomb is needed manned aircraft are still needed to carry them. This will probably change in the future, but is correct for several years at least.

The effect of drones on the fighting in Nagorno-Karabakh was a replication of events in Syria and Libya. Though again, one should be wary of statements over-hyping the effects there too.

“Turkey used its fleet of drones to lay waste to Syrian Arab Army (SAA) tanks, vehicles, and air defenses, while Azerbaijan was able to do much the same against Armenian forces in Nagorno Karabakh.”[xxi]

This misrepresents events in Nagorno-Karabakh and in Syria. Syrian regime forces were initially surprised and shocked as they had no air defence assets facing a Rebel force devoid of aircraft. However, after suffering many fewer casualties than claimed in press releases by the Turks, the Syrian recovered within 24 to 48 hours, halted the Rebel ground attack the Turkish drone-offensive was supporting and counter-attacked to retake all ground lost to the Rebels and more. In Syria and Libya, the drones attacked in a permissive environment regarding anti-drone defences. However, it should be remembered that manned aircraft have been operating like this for many years, even when the enemy ostensibly has some air defence capability – see the Israeli air force’s almost complete freedom of action since 1982.

However, exaggerated though the hype may be, the obvious lesson from all these events is that ground forces need to invest significantly in developing and procuring effective anti-small-drone equipment. Once the technological issue is solved, as it should be fairly easily, the tactical issue must be addressed – training units to deploy and maneuver together with the new equipment so as not to accidentally move outside the protective dome they provide and learning to operate one’s own drones through that dome.

Drones do provide some new tactical capabilities: longer loitering times compared to manned-aircraft; the ability of the pilot, sitting in an office, to calmly survey the ground and focus on detection and targeting and when he tries – to exchange seats with someone fresh; the quality of the pilot’s surveillance equipment. But they do not, in as of themselves, radically change the ability or utility of airpower on the battlefield. The results achieved in Nagorno-Karabakh were not better than those achieved by drone-less properly handled air forces in previous wars. Had the Azeris employed an air force with capabilities similar to those of the USA, Israel or similar armies, the result would have been at least the same, and some would argue even better – given the more powerful bombs carried by manned aircraft.

A tactical revolution is not in the offing, however a strategic revolution is. It comes not from the tactical capabilities of the drones, but from their cheapness, simplicity and availability compared to manned aircraft. States and organizations who cannot afford a full-capability air force of manned-aircraft can now acquire a capability that may not be as comprehensive or as powerful as manned aircraft. Thus this is a huge leap from nothing, or almost nothing, to capabilities they could only dream of. For states like the USA, Western Europe, Turkey and Israel, with large, advanced air forces of manned-aircraft, the drones are an incremental, albeit useful, improvement. For states like Azerbaijan, unable to fund and maintain an air force, though it had a smaller weaker air force, this was a radical enhancement in military capability. For an organization like Hezbollah, which cannot even establish and maintain an air force like Azerbaijan’s and which only began to use armed-drones during its involvement in the Syrian Civil War, this is an enormous leap up. For decades the Israeli army has been used to fighting without looking up to see whose aircraft was rumbling overhead, knowing with virtually 100% certainty it was Israeli. It can no longer be certain of that and must prepare to operate under unfriendly skies. Achieving air-superiority in one fell swoop as in 1967 is no longer an option. That is undoubtedly true also in other parts of the world.

Therefore, the lessons of Nagorno-Karabakh are that advanced air forces have very little to learn from this war. Conversely – air defence forces and ground forces, even of armies that have advanced air forces, must take into account and prepare to meet a new threat that enables poorer and even primitive military forces to create an aerial threat that did not exist before.


[i] Uzi Rubin, The Second Nagorno-Karabakh War: A Milestone in Military Affairs, The Begin-Sadat Center for Strategic Studies, December 2020,, pp 4, 5. The quotes from this source are only an example of many similar claims.
[ii] The Azeris do not admit this but comparing their statements at different times reveals their inaccuracy. On the Armenian side, nine days after the war ended, Armenian General Movses Hakobyan, Chief Military Inspector of the Armed Forces, resigned and claimed that all the official statements by the Armenian Ministry of Defence had been lies. "Movses Hakobyan reveals military secrets about Armenia's defeat", Vestnik Kavkaza, 19 November 2020, Even without accepting Hakobyan's statement as completely accurate, a detailed reading of those statements suggests a great deal of obfuscation.
[iii] Officially the Armenian side included two separate political and military entities: the Republic of Artsakh in the Nagorno-Karabakh territory and its ally the Republic of Armenia which only assisted Artsakh. For this article, the difference does not matter. For the sake of brevity and clarity and without going into political and legal implications, they will be referred to in general as Armenians.
[iv] Stijn Mitzer, Joost Oliemans, Jakub Janovsky, 'Dan' & 'COIN', "The Fight For Nagorno-Karabakh: Documenting Losses on The Sides Of Armenia and Azerbaijan", ORYX, The list is updated occasionally as new information is found. The version downloaded is updated to 25 October 2021. The current author has considered and corrected a couple of small errors.
[v] President Aliyev's Speech to the Nation, 1st December 2020,
[vi] Aliyev's statement did not include the wide variety of other utility vehicles – jeeps, vans etc. According to Oryx data, these include approximately 85 more vehicles destroyed by drones.
[vii] P. H. Pukhov (Editor), Storm Over Caucasus (Russian), Center for Analysis of Strategies and Technologies, 2021, p 49.
[viii] Here too propaganda issues muddy the waters – the Azeris deny, though multiple other sources attest to, the involvement of Syrian mercenaries, who apparently suffered 250 to 540 killed depending on the source. Also, the Azeris have not released total figures for wounded – only that on 3 December 2020, three weeks after the war ended and about 10 weeks from its beginning, 1,245 soldiers were still in medical institutions (Azerbaijan Ministry of Defence Statement, 3 December 2020). Given the usual statistics of wounded versus killed in wars conducted with large forces employing mostly heavy weapons, typical severity of wounds and recovery times that number suggests several thousand wounded throughout the war. The Armenians claimed to have inflicted 7,630 casualties (Armenian Ministry of Defence Spokesperson, 8 November 2020) . Their statements were often proven inaccurate so should be used with caution – this would suggest about 4,500 wounded.
[ix] Official statement by the Armenian Ministry of Defence, 15:00, 8 November, 2020. This was their last statement on the subject though the war continued for another 39 hours. Photographic evidence collected by the Oryx team documents the destruction of 91 Azeri tanks, infantry fighting vehicles and armoured personnel carriers, another 26 damaged and a few more abandoned and captured by the Armenians. However, the Armenians had much less photographic evidence to provide, because drones photograph their actions whereas ground weapons do not.
[x] Official Azeri statements claimed successes on all these days with lists of Azeri villages abandoned in the previous war (1989 – 1994) were liberated. Even assuming they are telling the truth, the locations of these villages shows they are all right on the front line, a bit to the right or a bit to the left of the previous location claimed to be liberated. Also, as the war progressed, journalists visiting 'liberated' villages often found them still in Armenian hands. The Azeris apparently conducted a number of photo-op raids in which combat teams with cameras infiltrated behind Armenian forces to film the liberation of locations but withdrew shortly after.
[xi] There have been reports of Armenian troops and units deserting against fewer reports claiming the same on Azeri troops. In war, there are often moments of crisis in which men and units break. Sometimes these are permanent and sometimes temporary. It is possible that in the last days of the war, more Armenians despaired so there was a greater inclination to retreat than to fight, but definitely during the first half of the war, the Armenians stood their ground and counter-attacked frequently and on the last days too there were fierce battles.
[xii] Official statement by the Armenian Ministry of Defence, 15:00, 8 November, 2020.
[xiii] "The bomber will always get through" – in a speech to parliament about the futility of trying to defend against the strategic bombing of a country's civilian heartland. The only way to win, suggested Baldwin, was to bomb the rival even more powerfully than he could bomb Britain so that the rival would capitulate before Britain was forced to.
[xiv] For an explanation of the technical issues of the equipment used by the Armenians, see an article written by a former Serb air defence officer who had operated similar systems: Zoran Vukosavljević, "Okršaji azerbejdžanskih dronova i Sistema PVO Jermenije: Koje su lekcije za Srbiju?", Tango Six, 20 Oktober 2020,
[xv] Optimization to see targets only bigger than… and faster than… is done to reduce false detections from various phenomena including birds.
[xvi] Ashot Hakobyan, "General Daniel Balayan: 'Azerbaijanis brought this fight upon themselves ",Aravot – Armenia News, 21 July 2020,
[xvii] "Russia Shot-Down A Total Of Nine Turkish Bayraktar Drones Near Its Armenian Military Base – Russian Media Reports", EurAsian Times, October 21, 2020,
[xviii] " Armenia: General accused Prime Minister Pashinyan of blunders in the war in Karabakh" (Russian), BBC News – Russian, 19 November 2020,
[xix] See for example the shooting-down of an American surveillance drone by Iranian air defences in summer 2019.
[xx] Erwin Rommel, The Rommel Papers, De Capo Press, 1953, p 285.
[xxi] John Flannelly, "Drone Effectiveness Against Air Defenses, Not Tanks, Is the Real Concern", The Defence Post, 7 December 2020,