How to Recognize Cyclical Patterns Before They Blow Up
I spent most of 2023 watching people rebuild the exact same mistakes from 2021, which is not a comfortable place to be when you have been in finance long enough to remember where this script starts. The thing nobody tells you about spotting repeating history is that the patterns are not clever. They look different on the surface but the mechanics underneath are almost identical, and the reason they repeat is not ignorance so much as it is a structural forgetting that happens every generation or so. 2021 is probably the cleanest single year to study right now because almost every major speculative cycle ran through it in rapid succession, compressed into twelve months instead of decades. That compression makes the repetition much easier to track than it was in prior eras. I keep a simple spreadsheet where I log the signals I watch, and the 2021 entries were the longest I have ever had before a cycle resolved. Let me walk through how I actually track this stuff in practice, because the theory is easy and the practice is where people mess it up.
The first thing I look for is the velocity of adoption relative to utility. In 2021 you saw meme stocks and NFTs move on pure social velocity while the underlying economic value stayed basically flat. I noticed this in January when GameStop moved on retail coordination rather than earnings or cash flow, and then again in March when crypto infrastructure projects started launching with zero revenue models and still raised hundreds of millions. The repeat was not subtle. It was right there. I also watch leverage structure closely. In 2021, DeFi lending protocols like Aave and Compound ran with utilization rates above 90 percent for months while assuming collateral values would never correlate downward simultaneously. That is the same assumption that killed Long-Term Capital Management in 1998 and set up the subprime crisis in 2007. The wrapper changed. The math did not. When I saw Treasuries and risk assets move together in March 2020, I marked that as an early warning signal because it meant liquidity was about to become the primary driver instead of fundamentals. That signal repeated in February 2021 when the Federal Reserve switched to outright quantitative easing and everyone immediately re-priced risk upward. I logged it the same way I would have logged it in 2009 or 2015.
One specific edge case I ran into during 2021 was tracking stablecoin depegs. Most people only paid attention to Terra/Luna in May, but the earlier depeg events in Algorithmic USDD and various renminni-pegged tokens in late 2020 and mid-2021 were the real test. I learned to watch the on-chain collateral ratios of algorithms in real time rather than waiting for the price action to confirm the model was breaking. Waiting for the confirmation always means you are late. I built a simple alert that triggers when any algorithmic stablecoin drops below 1.05x backed reserves, and that gave me roughly forty-eight hours of advance notice on the major moves. It is not perfect but it is better than nothing. Here are the clearest repeat cycles from that year and the years around it.
Get the Full Details
The Dot-Com Script, Replayed With Better Fonts
The year 2021 had a version of the dot-com bubble playing out in real time, and the structure was nearly identical even though the players did not know the old one well enough to avoid it. The key similarity is not the stock picks, it is the capital formation pattern. You get cheap money, a new distribution channel, venture backing flowing in faster than due diligence can catch up, and then a cluster of companies that can justify their valuations only if growth continues at the current rate forever. In 1999 that was internet infrastructure. In 2021 it was everything from SPACs to NFT marketplaces to DAOs. The SPAC mechanism alone produced over six hundred deals that year, many with no revenue and no clear path to profitability, structured so that sponsors walked away with equity regardless of whether the public shareholders did. That is the same structural incentive that existed in pre-1929 investment trusts and again in 1999 with the dot-com rollups. The counter-intuitive part that most people miss is that the repeat does not require everyone to be irrational at once. It only requires the marginal buyer to be irrational, and the marginal buyer in 2021 was usually someone entering the market for the first time because social media made it look easy. I watched this happen with NFTs in the spring of 2021 when OpenSea volume hit approximately three billion dollars in a single month and the average buyer had never held a collectible asset before. The repeat was happening, but it was being read as novelty rather than as a cycle.
Leverage Cycles That Feel New Every Time
The 2021 crypto leverage explosion looked completely different from 2008 because it happened on-chain instead of in mortgage securitizations, but the underlying dynamic was the same. You had leveraged positions funded by rising collateral values, which required ever-rising prices to maintain, which required new capital to enter, which required optimism, which required more capital inflows. The important detail that gets missed is the deleveraging mechanic. In 2008, deleveraging happened through bank balance sheet write-downs and forced selling of illiquid assets. In 2021, deleveraging happened through automated liquidations on trading platforms, which created cascading fire sales because the liquidations were all triggered within minutes of each other instead of spread out over months. This is why crypto crashes feel faster and sharper than traditional market crashes. The automation removed the human delay that used to give markets a few hours of warning. I also tracked the private credit market growing from about two trillion dollars in 2020 to over three trillion by late 2021, with many of those deals structured with floating rate debt and covenant-lite terms. That is the same structural feature that amplified the 2007-2008 collapse. When rates rise and collateral values fall, covenant-lite loans provide almost no protection to lenders, and the lenders then have to sell aggressively into falling markets, which pushes values down further. It is a feedback loop, not a one-time event, and recognizing the loop early is what separates people who exit on time from people who hold through the unwind.
Housing and the Illusion of Permanent Appreciation
US home prices rose roughly forty-six percent between March 2020 and mid-2022, which felt unprecedented to a lot of people but was actually just a replay of the post-2012 recovery combined with an artificial demand shock from pandemic-era monetary policy. The repeat was in the financing structure, not the price direction. In 2005, the key signal was the ratio of home prices to median income crossing 5.5x and staying there while adjustable-rate mortgages dominated new originations. In 2021, the ratio crossed 6x but the mechanism was different because rates were near zero and refinancing was trivial. The danger was the same. People were buying at the top of a cycle using debt they could barely service if rates moved even slightly higher, and the assumption was that prices would only go up. The specific signal I watched in 2021 was the shrinkage of the first-time homebuyer share to below twelve percent, which is the same floor it hit before the 2006 collapse. When first-time buyers drop out of the market, you are left with investors and move-up buyers, which means demand becomes speculative rather than utility-based. That is the point where the cycle starts turning even if prices keep climbing for another six to eighteen months.
I also monitored the rent-to-price ratio in major metros. In 2021, the annual rent on a median home in most large cities was less than three percent of the home price. A three percent cap rate implies a seventy-year payback period, which means the entire expected return comes from appreciation. That is a speculative structure, not an income structure, and it is the same one that existed in 2005 across most of the Sun Belt.
What Actually Works for Tracking These Patterns
I keep a small set of metrics that I check weekly and I update them manually because automated feeds tend to lag during the critical window. The metrics I actually use are M2 money supply changes, the yield curve inversion spread, total margin debt, crypto total market cap relative to global equities, and the percentage of IPOs that are unprofitable tech or SPACs. None of these are individually predictive, but when three or more of them start moving in the same direction at the same time, the probability of a cycle inflection increases significantly. In 2021, margin debt hit records, the yield curve inverted in May, crypto market cap went from roughly eight hundred billion to over two trillion, and unprofitable issuers dominated both the IPO and SPAC markets. All four signals were firing at once, which should have been the most obvious warning possible. The fact that it was not obvious to most participants is the main reason history repeats, not because the data is hidden but because the narrative of the moment is too loud to hear the data. I also recommend tracking the time between peak leverage and peak sentiment, because in every major cycle since the 1980s, leverage peaks before sentiment does, and the gap between them is usually three to eight months. In 2021, leverage peaked in crypto around April-May and sentiment peaked around August-September. That gap gave people who were paying attention a window to reduce exposure. Most people were still adding exposure during that window because the headlines were still positive.
Where This Approach Fails
I need to be honest about the limitations here because this method is not a crystal ball. The biggest failure mode is when the central bank decides to absorb losses directly, which happened in 2020 and 2021 with the Fed's corporate credit facilities and swap lines. When the government backsstop a market, the usual deleveraging mechanism breaks temporarily, and cycles extend longer than historical patterns would predict. This happened in 2020 and again in 2021 with certain segments of the credit market, and it means any purely historical model will underestimate the duration of the peak. Another failure mode is structural change. The 2021 NFT and crypto cycles included features that did not exist in previous cycles, such as programmable ownership and composable DeFi protocols. These features can create new types of mania that historical patterns do not capture well. I have seen people try to force 2000 dot-com frameworks onto 2021 NFT dynamics and get the timing wrong by months because the mechanics of how value flows through those systems are different even if the psychology is the same. The most practical workaround is to treat historical patterns as directional indicators rather than timing tools. They tell you whether you are in a speculative environment and whether leverage is elevated, but they will not tell you the exact month a cycle ends. If you need precise timing, you should supplement the pattern work with on-chain data for crypto, quarterly margin debt reports for equities, and monthly housing starts plus refinancing activity for real estate. Those harder data points will flag turns before the broader cycle metrics do.
Why This Still Matters in Practice
I keep coming back to 2021 because it was the last year where every major speculative cycle overlapped in a way that made the repetition impossible to ignore if you were looking for it, and also the last year where ignoring it was still the rational choice for most people because the narrative was so strong. The reason history repeats is not that people are stupid, it is that the incentives to participate in a bubble are almost always stronger than the incentives to sit it out, and the penalties for being early are financial and social while the penalties for being late are mostly regret. If you want to track this yourself, start with the four to five metrics I listed above and log them weekly for six months. You will start seeing the overlaps between signals without needing a PhD in macroeconomics. The spreadsheet is not going to make you rich, but it will keep you from being surprised when the next version of the same story comes around, and that version is already writing itself in the data.
