Ukraine is winning. How much they win is up to us.
I have short but successful track record predicting the future on Metaculus and I’m in the 4.4th percentile in the Ukraine tournament (131/2970).
Last year I was among the first to predict the war would directly cause 100,000 deaths, I joined the “Russia won’t capture any of the major cities” bandwagon pretty quick, I was bullish on the fall of Mariupol (though Ukrainians did a great job holding out) and I expected Ukraine to retake Kherson (though the timeline was fiendishly hard to predict). I expected Putin would stay in power (>80% chance) with a high approval rating.
Finally, I predicted on May 13 that Russia would control less territory one year later (88%) —which is clearly going to happen, with not one but two large regions returned to Ukraine:
So, what will happen in 2023? After nearly a year of war I’ve still found no good sources to reliably estimate key statistics such as personnel losses on each side, or the extent of corruption on each side. Meanwhile, most commentators I’m seeing have only average-to-decent epistemics (so, not great). Perun is great, but is unsure himself which sources to trust. In early 2023, I found events harder to predict, as it was getting hard to tell who had the advantage.
However, Russian failures in its recent offensive imply that Ukraine has better strategy and tactics, even as Russia has strained its supply of equipment; the question is whether Ukraine’s advantage is big enough.
Ukraine’s main advantage is always western military aid (though their will to fight is also high, bolstered as it is by Russian kidnappings of children, “Ukraine doesn’t exist” rhetoric, missile terrorism and the killing of so many soldiers who wanted to be civilians). Countries near Russia such as Estonia, Latvia and Poland have the strongest support for Ukraine:
But their military might just isn’t very big, so support from the US, EU, UK and Germany are more decisive.
The war hurts the global economy, but mainly not due to military aid or refugees. Indeed, such things need not hurt GDP at all: just as vandalized windows lead to increased window production, war can be good for business even as it harms our quality of life. The war’s effect on global GDP is still negative, though, and much greater than the cost of military aid or refugees:
As America buys new hardware they can always send their old stuff to Ukraine, and that stuff is superior to most of the Russian stuff. Problem is, the U.S. just isn’t sending that much of its old stuff. For example, the U.S. has thousands of Abrams tanks but only plans to send 31. And those awful cluster bombs Russia was dumping on Ukraine early in the war? For some reason the U.S. built a huge stockpile of 5.5 million of them (with ~730 million submunitions). Ukraine’s begging to get some of those with no luck yet. More generally, the U.S. has huge stockpiles of old munitions awaiting “demilitarization”. Presumably the U.S. could save money on demilitarization by sending that stuff to Ukraine instead. Overall, U.S. military spending on Ukraine less than 6% of its annual military budget.
Corruption, incompetent commanders and soviet-style thinking aren’t limited to the Russian side, either.
Even so, there are reasons to be optimistic. For one thing, Russia’s military stockpiles remain large, but not compared to NATO’s. Just a few tens of billions more in aid should push Ukraine firmly in the winning direction (again, this may sound like a lot but it’s a tiny fraction of the military budget). And let’s review recent Russian advances near Bakhmut:
North of Bakhmut and Soledar, Russia didn’t fare so well. Overall, Russia lost about 6% of Ukraine since a year ago — tens of thousands of square kilometers.
Russia has consistently lost tanks at more than triple the rate of Ukraine…
…and since Ukraine has captured large numbers of tanks and other equipment in certain times and places…
…it looks superficially like the Ukrainian military gained hardware since the invasion started. In reality there has been a lot of attrition and hidden losses, not just from equipment being destroyed but also from ordinary wear and tear; war is very hard on military equipment. But western aid continues and is likely to bring some further success.
Here are some predictions for 2023 that I’m fairly confident about:
- Russia won’t capture any major Ukrainian city this year with population over 500,000 (99%)
- Russia won’t invade Finland or the Baltic states (99%)
- NATO still won’t put a no-fly zone in place (98%)
- Belarus won’t launch an invasion of Ukraine (97%) because neither Lukashenko nor Belarusians want to, and Putin lacks leverage
- Russian troops won’t launch a major offensive from Belarus (90%) due to a lack of manpower and equipment
- In the last 9 months of 2023, Russia will fire fewer missiles at Ukraine from Russia/Crimea than they did in the first month of their invasion, i.e. 1200 or less (75%)
- Ukraine will launch an offensive this year (96%) in the south (80%). The exact path of the southern offensive is probably not known even by Ukraine’s military, as it depends on recon results.
- A Ukraine offensive will advance further than the Russian offensive near Bakhmut did, i.e. more than ~10 km (80%), while taking fewer casualties (80%) although this last point may be hard/impossible to assess. It looks like a southern offensive is starting already but it could be a feint or recon. Public information suggests Ukraine doesn’t have enough equipment yet (or even enough trained soldiers) but Ukraine’s allies could have delivered more equipment secretly and/or Ukraine may take the risk of using up old equipment before new equipment arrives. The biggest reason to think this is a real offensive push is that a new batch of Ukrainian soldiers just returned from training in Britain. I hope it’s not starting yet, as Ukraine probably isn’t ready.
- Putin will stay in power (91%); the #WindOfChange isn’t blowing strong
- Russia won’t nuke Ukraine or the West (98%)
- Contrary to Russian claims, Ukraine will not capture any Russian territory beyond 5 km from the Russian border (95%); stunts by Russian militants don’t count.
- Ukraine won’t capture more than 0.2% of Crimea this year (80%) and certainly less than 2% (95%).
- Ukraine won’t capture any territory that Russia/“separatists” already held before Feb 2022 (95%) and Ukraine won’t launch a large offensive in the east (75%) unless early probing attacks go unexpectedly well.
- Russia won’t hold any Ukrainian territory more than 30 km beyond current front lines by the end of 2023 (75%).
- Russia will not capture the entire Donetsk oblast (90%) and they probably won’t even recapture Luhank (80%).
- Ukraine is generally taking fewer casualties than Russia (98%), even now (80%). It won’t be possible to verify this.
- For fun, here are my probabilities on former Russian PM Medvedev’s silly 2023 predictions: (1) 8% (2) 1% (3) 0.1% (4) 0.0001% (5) 0.0001% (6) 0.00005% (7) 5% (8) 0% (9) 0.5% (10) 0.01%. In other words, Medvedev is full of sh*t and the chance that even one of his predictions comes true is about 10%.
Now for some more uncertain estimates:
- Bakhmut will fall by the end of May: 50%. It is confusing that Ukraine hasn’t abandoned Bakhmut because the Ukraine-Russian casualty ratio is probably higher than it would be if Ukraine pulled back. The deputy defense minister said it’s a strategic decision, but I do find it a bit hard to understand. Anders Puck Nielsen interprets this optimistically as a perverse kind of risk aversion — the risk being Russia taking a pause. Edit: another reason is to protect Sloviansk and Kramatorsk from attack by keeping them out of artillery range. It’s better for Russia to get one utterly destroyed city, rather than Russia getting one mostly-destroyed city and Ukraine having two more partly-destroyed cities.
- Avdiivka will fall by end-of-year: 40%
- Even money says Ukraine cuts the land bridge to Crimea (50%), but conditional on doing so, they have a good chance of also retaking the Zaporizhzhia nuclear plant within a year (80%). Ukraine is limited by equipment and training, on the one hand (as many experienced men have died), and freshly-built Russian defenses on the other. What makes success conceivable at all is the incompetence of Russian military leaders and the limited skills of the Russian mobilized.
- Ukraine will get older fighter jets from somewhere: 65%
- Ukraine will win back a substantial amount of territory this year: 65%
- Ukraine will advance more than 25km into Russian-held territory: 65% (and likely above 10km by mid-May: 50%)
- Conditional on the southern Ukrainian offensive failing to reach 40% of the way to the Azov sea by July, they likely won’t cut the land bridge by the end of 2024 either (60%), mainly because the West is likely not to send enough weapons and Russia will gain the upper hand. On the other hand, conditional on cutting the land bridge, Ukraine is likely to strike one or both Kerch bridges again (70%) but is unlikely to capture 10% of Crimea within a year afterward (75%) or even within two years (65%).
- Investigators may announce that Russia attacked Nord Stream 1 (55%), but if so, Putin’s boot-lickers will continue believing Hersh’s nonsense. But Putin actually did attack Nord Stream 1 at Russia’s expense (95%) though not Nord Stream 2, as Putin likely hoped to use NS2 to ship Russian gas (for a propaganda victory against the U.S.’s pledge to stop NS2). U.S. and Norwegian investigators probably already knew last year what had happened, but deliberately said nothing. Note that Nord Stream 2 is still usable.
Edit: As Ukraine’s counteroffensive begins (June 5 AM), I’m updating some probabilities:
- Ukraine advances more than 25km into Russian-held (Ukrainian) territory in 2023: 85%.
- Over 50km: 75%
- Ukraine cuts the land bridge to Crimea: 70%
- UKR wins back substantial territory: 75%
- Putin will stay in power this year: 75%. I also give a 25% chance that he stays in power as a sort of figurehead who is hardly pulling the strings anymore, but if this happens it will probably not be clear. One curious sign: it looks like Prigozhin committed treason but Putin has kept mum about it. I have low confidence in this prediction.
- Avdiivka will fall by end-of-year: 30%
June 29: Ukraine isn’t doing as well as hoped. Reducing probabilities:
- Ukraine advances more than 25km/50km into Russian-held (Ukrainian) territory in 2023: 75%/65%
- Ukraine cuts the land bridge to Crimea: 60%
- Ukraine wins back substantial territory: 65%
July 18: weeks after major attacks on Russian logistics (Rykove ammo depot and Chonhar bridge) Ukraine didn’t follow through, but it may just be that the attritional phase is successful but slow-going. It’s encouraging that Russia is demoting/arresting various commanders, but I’m still reducing probabilities:
- Ukraine advances more than 25km/50km in 2023: 60%/45%.
- Ukraine cuts the land bridge to Crimea: 35%
- Ukraine wins back substantial territory: 40%
July 21: 25% that Ukraine cuts the land bridge
Aug 7: 15% that Ukraine cuts the land bridge; 60%/35% for 25km/50km.
Sept 15: 10% that Ukraine cuts the land bridge; 35%/20% for 25km/50km. It’s been disappointing to watch this play out. Ukraine has been doing its best to impede Russian supply lines and destroy Russia’s air defense, but it has been impeded by the limited amount of equipment it has received from the west, while Russia has shown surprisingly strong military capacity. Ukraine may well reach the third major defense line south of Verbove, but it would be mildly surprising if they even reached the north side of Tokmak (17 km from the current front line and about 24 km from the original front line). Most likely, they’ll reach roughly 14 km in from the original front lines, in a single location, by end of year. While Ukraine occasionally penetrates a defensive line, they have so far not pushed any tanks through antitank obstacles, and even if they did, progress is so slow that Russia will have time to build new defensive lines. So the only ways to get a breakthrough at this point are (i) to exploit a weak spot very quickly (e.g. via the Dnipro river, which is now very narrow after Russia caused the Kakhovka dam to collapse catastophically), (ii) to hinder logistics on a scale they never managed to achieve before (probably by blowing up key bridges), or (iii) to destroy so much Russian stuff that RuAF faces a serious shortage that compromises their fighting power. It’s unlikely that any of this will happen soon. Still, it’s even more unlikely that Russia will make larger gains than Ukraine by end-of-year ― even though Ukrainian lines appear to be weaker. The most likely outcome this year is a stalemate, although RuAF seems likely to face some kind of shortage a few months from now, which Ukraine can then exploit if the west provides enough support. Ukraine likely could’ve cut the land bridge this year with more western support delivered faster. Sadly, support was somewhat tepid, e.g. with the U.S. pledging zero fighter jets and only 31 tanks, which still haven’t appeared on the battlefield. It looks like support will stay healthy for at least the next year, but Putin is in it for the long haul, and matching long-term commitments from the allies are needed to counter that.
For more predictions, visit the Ukraine tournament on Metaculus.
One issue I expect to arise with these predictions is that they probably won’t appear well-calibrated because outcomes will be correlated with each other, ultimately creating an appearance of underconfidence or overconfidence. Historically, my bias has been mild underconfidence.
And by the way, it’s hard not to look at the difficulty of Russia’s defensive lines and wonder where the Boring Company is when you need it. I’m gonna give a 10% chance of a big tunnel, just so when they burst through the second defense line I can be the one guy saying “I predicted they might do that.”