That's... not what front running is. Front running is a problem of agency relationships. The classic example is you placing an order with your broker, and your broker using that information to trade ahead of you. Your broker is your agent and has a duty not to carry out your requests in ways adverse to your interests. What this article (and Flash Boys, for that matter, which is a terrible book) is describing is something more like price discovery.
I think you are arguing semantics. The general understanding of front running is that you intend to make a trade, and expect it to execute without anyone knowing before execution that you wanted to do it. Any activity made by others based on the knowledge of your trade before execution that affects your trade can be put in the bucket of front running.
In blockchain, the miners control the order of transactions. They do all manner of things to extract profit from your trade because of that control. The simple solution, which is good enough for most people, is to restrict the slippage and assume the miners will get as close to that number as possible. But people are thinking of this problem very hard and ways to solve it.
That seems like a crazy definition, since fully realized it expresses the idea that people should be able to make large trades atomically without moving markets.
I don’t think that’s crazy. I think most people think it’s fair that the plumbing between their decision and the action shouldn’t be used to profit from them. Of course it happens in tons of ways. But that’s the general feeling. They want to see an order book, and if the book quotes X price, the seller won’t change it when they click the Buy button, etc.
It's not about doing trades without moving market afterwards, it's about market moving _before trade_ as a preliminary response of the trade existence.
But trades aren't atomic, even though the i-bank traders in Flash Boys wanted to get paid to pretend they were. If you're trying to unload N shares, that's going to happen in some number of k<=n-sized chunks.
I'm not sure how this can be applied to the discussed issue? Especially considering that this kind of trades are atomic (when run by miners/flashbots), though maybe we use different notion of "atomic".
I think that is too simplistic. For example, in a CLOB market putting something in the order book is visible and depending on price and volume will not just execute in one go and therefore cause change. But it is a market reaction to a published intend to trade.
Issues around market microstructure, liquidity etc are very well researched.
You do not understand me. Reactions to an order make sense. But if I click "Buy", and before the order hits the book, the software vendor that is processing my order notifies a market maker to increase their price in order to make me pay slightly more, this would be unfair ("front running").
The miner is your broker because the miner is executing the uniswap smart contract and it can artificially delay your transactions and insert its own transactions.
However, it's not front running legally, because the transactions are public.
AFAIK, much of what they call Rejected Energy in that LLNL illustration is lost as heat in the ways you'd probably expect - for example, an internal combustion engine gets hot, or a gas-fueled power station gets hot
I'm not really sure what the author is trying to communicate here. Cryptocurrency is volatile? Who knew! Of course that volatility comes from somewhere, there is no true digital entropy. Anyone who can chase down the root cause can profit off it, that's the nature of most trading anyways.
His energy consumption argument seems to be false at a glance. I don't claim to be an expert but I can do some basic addition at least. If you have two components in consumption - "constant" and "variable", and if you have two components in production - "used" and "wasted", then adding strictly to the "constant" part of the consumption you also add strictly to the "used" part of the production. Constant energy receiver doesn't affect variable/spike part of production and consumption. To me it's an obvious conclusion, but of course being an amateur I may be wrong here.
I think if you talk about the behaviour of the system without talking about the technology it becomes quite clear that this system is just poison. You have to pubish your deals before they execute, and then everyone else has an opportunity to bribe the people processing the deal to screw you up. Who actually wants that behaviour?
How is this different from what we have today with payment for order flow and HFT? I’m no crypto bull, but I kinda prefer a natural and transparent system to a tightly regulated one steeped in corruption.
So there are 2 things there. There's payment for order flow - Robinhood will route your order to Citadel and Citadel will choose whether to trade with you at that price (or route your order somewhere else), the big difference though, is that Citadel is legally obliged to give you the NBBO (national best bid offer) so whilst Citadel is paying to be the counter party on your trade, it doesn't change the deal you get. It's no worse for you - you still get the best price, but there are various reasons Citadel wants to take the counter-side of that trade.
HFTs are again, not able to effect your trade. If you place an order in an exchange then HFT's will respond once you have traded, but your trade is guaranteed to go into the market and execute without anyone interfering before hand. If there was liquidity in the market when you tried to trade at the price you entered, you will get that trade. Now it may be that an HFT sees your trade in Chicago and goes off and trades on some other exchange super quickly based on the new information or they might cancel their remaining orders in the makret, but they cannot interfere with your trade before execution.
What's unique about the ethereum situation is that different actors int he market can interfere with your trade before it executes because your desire to trade gets published before it's executed and the miners (in this case I guess sort of analogous to the exchange) are willing to prevent you accessing the market due to a bribe from a third party based on the information you were forced to publish. Which means you never have confidence that you can actually execute a transaction you beleive is available in the market.
For one, traditional finance has the NBBO requirement. Regardless of how an order makes its way to to a broker, they can’t execute at a worse price than is available.
Payment for order flow (at least in the case of retail brokerages) is more about the big players preference for trading with unsophisticated players. The rubes just lose more often without any malfeasance required on the side of the brokerage
The author is absolutely wrong on power grids. The power grid is over-provisioned in the sense that it has a lot of power on stand by. The physical electrical grid is really balanced more or less instantaneously. Inertia in generators together with droop control take care of this balancing process. But you really do burn less fuel when there is less demand.
The energy is not wasted, it is ordered, not demanded and thus never produced. If it were produced it would actually be extremely difficult to get rid of this much energy. You can't just easily dump energy.
As the author is waving their credentials around, I am an active researcher on renewable power grids. It's a bit scary to think that someone with such a basic misunderstanding of the power grid was asked to referee a paper on this, but then a lot of marginal research is published everyday and referees have to come from somewhere...
There is no rejected energy. If there is an overproduction of energy, the price goes down (it may even go negative), and industries will consume the energy.
"Rejected" does not mean that the energy gets sent back to the producer like a parcel.
"For a coal fired power station, for instance, about 2/3 of the energy released when the coal is burnt is discarded as heat in the environment. This reject energy sometimes appears as clouds of vapour coming off a power-station’s cooling towers, such as the well-known ones at Didcot in England."
This is a complete misinterpretation of the graphic. The "Rejected Energy" are the losses that occur when converting primary energy on the left into useful energy on the right. In a car, for example, this would be the difference between the chemical energy contained in the gasoline and the mechanical energy used to accelerate the car and overcome the drag during driving. In a thermal power plant it would be the difference between the heat released when burning the fuel and the electricity produced.
As others have pointed out, the power grid does not even have the capability to "burn off" excess power. It is simply nerver produced in the first place.
The graph is absolutely meaningless to me without further contex. So, the electric grid generates 12.7 "units". If we subtract solar, wind and hydro then we get down to 6.42 "units". I have chosen these because they are not thermal power sources. These power sources generate an almost equivalent amount of primary energy and electricity. So we are left with 6.42. 6.42/31 is 20%. Where are these 20% efficient power plants? Nuclear alone would be 30% efficient. Combined cycle gas is 50%. The numbers make no sense to me.
Edit: Ok I read the note at the bottom and the graph literally adjusts the rejected energy of renewables to match thermal power plants. Assuming 33% efficiency that means 12.56 phantom units of energy that never even existed in the first place are being rejected.
Additionally assuming this power is rejected through shunts (effectively resistor banks) this power is from peaking plants shutting down or over provisioned baseload thermal plants.
The only way Bitcoin would use this energy is if they had a contract with the power company to be an on-demand load, switching on and off to help smooth the peaks.
If they aren’t they’re just driving up power use and forcing more operation of less efficient reserve generation on peak.
I understand that a lot of the servers are close to large hydro plants. Under-utilised hydro plants can simply spill water - they spin up and down rapidly compared to thermal plants.
> The only way Bitcoin would use this energy is if they had a contract with the power company to be an on-demand load, switching on and off to help smooth the peaks.
But that is exactly how power plants are using Bitcoin. That's the idea behind Seetee.
Isn't that a waste of the capital used for the Bitcoin mining hardware itself? Why would a miner shut down their profitable operation simply because a coal plant turned on?
The miner and the power plant operator are often the same entity. They turn the mining equipment on when they want to dump energy, and off when they have more productive places to sell it. Seetee is a good example.
If you own the mining hardware, why would you do that? You are better off running that hardware 100% of the time (using batteries or similar when necessary). Mining hardware is expensive.
Only when you are actually better off. isn’t there a crossover point between the outside market needing/paying for electricity and mining? Or is mining the most effective return in all cases?
Animats explained the meaning of rejected energy in the diagram below. It's efficiency losses in the entire system (I wasn't aware of this terminology). The actual size of load balancing energy used is quite small.
I didn't know about Seetee but potentially there are some minor opportunities to mine bitcoins (effectively as an on-demand load) to help balance the grid at short time scales. It also makes sense for a Norwegian Energy company to do this. However this is a tiny amount, especially compared to the efficiency losses the author is presenting in the graphic. I looked up some data I had handy right now and I think we are looking at a few percent of the overall energy consumed at most.
- The visualization used by the author shows US energy consumption while ~65% of mining takes place in China[0] where burning coal remains the dominant energy source.[1]
I guess you have made a wrong conclusion: coal dominance in total does not conclude to coal dominance for mining. Because mining factories are very sensitive about the energy price, they are mostly built near locations where energy is cheap enough and will be wasted if not consumed locally by the mining or transmitted to other places in time.
Do you know if the graph in the article is accurate, and what is meant by "rejected" exactly?
What surprised me is that "Transportation" in that graph is supposed to "reject" over 22% of its energy, which seems odd to me. For electricity I get it, it's hard to store so if you have excess production there's not much you can do. If your wind turbine can't dump its energy anywhere it's just lost, usually. But for petroleum can't you just... not use it?
As such I wonder if "rejected" means all losses? In which case it's unclear how bitcoin would fit into that.
In general I think the author doesn't make a good enough case. He'd have to show not only that there's enough loss to power the bitcoin network, but also that bitcoin actually slots into these losses instead of adding to the demand.
They mean rejected heat. That's basic thermodynamics. See "Carnot efficiency". Maximum possible efficiency of a heat engine is
(Tin - Tout) / Tin.
Temperatures are from absolute zero. This applies to all forms of heat engines. Large power plants run steam around 840°K. Typical outside temperatures are around 280°K. So maximum possible efficiency for steam plants at achievable temperatures is around 66%.
Actual numbers for coal, oil, and nuclear plants run lower, maybe 40%. The most efficient natural gas plants come close, around 60%, by running at higher temperatures and avoiding conversion to steam.[1]
Ok, but if that's the article's argument it really makes no sense, no? How would the author propose harvesting these various losses to mine cryptocoins?
It's a really weird take if that's what the author is saying.
That's what the diagram copied from Lawrence Livermore National Labs says about rejected heat. What the article author is trying to say is something else entirely.
It's mostly these sentences (in relation to that diagram) that I really fail to understand:
>If Bitcoin used only wasted energy it would consume >2/3rds of the energy of the planet.
>This is why hand-wringing about Bitcoin’s energy use is so frustrating. Bitcoin is pointing at the moon and people are yelling at its finger. The self-congratulatory ignorance I see in this debate is unreal.
I don't get what it means or what point the author tries to make here.
The best I can guess is that the author is mixing up load balancing through reserve energy and efficiency losses. The amount that is reserved for load balancing is also not anywhere near the size they are talking about.
> I don’t generally cite my own credentials because I’d prefer you were persuaded by my arguments than by my resume, but for the record in grad school I actually peer reviewed academic papers about load balancing the electrical grid. I’m not talking out of my ass when I say that most energy is wasted. If Bitcoin used only wasted energy it would consume >2/3rds of the energy of the planet.
The wasted energy he is talking about here is mostly waste heat and other losses, it is not energy that is actually available for use. That makes the argument pretty silly, since increasing power consumption incurs the exact same losses.
How does so much energy from petroleum go into "transportation" and then into "rejected energy"? Is the small text at the bottom saying effectively that transportation is an inefficient use of petroleum energy? If so that makes it a weird argument to suggest that Bitcoin is in some way making use of that 22% wasted transportation contribution.
Won't Bitcoin mining show up as part of use used residential and commercial electricity?
And, wouldn't "less coal and natural gas burned" be more desirable than "using that to mine Bitcoin" and sneering about "handwringing"?
It’s a terrible point that the graphic doesn’t support in the slightest.
But to answer your question: IC engines are energetically inefficient - at best a large fraction of the input potential energy escapes as heat, and actual car engines are worse than ideal.
The graph is very confusing. Rejected energy is basically thermal loss, it's completely unavoidable without switching to EVs powered by renewables.
Transportation is inefficient because ICE engines need to run at lots of different RPM ranges. Hybrid cars have more efficient engines because the engine can run at the most efficient RPM. EVs only have charging and battery losses.
Fyi the Github link on this is much more interesting than the article’s overview of it [1]. What happened is that Ethermine, the largest Ethereum miner, turned on a piece of software this week that takes advantage of their position as a miner to order transactions to their advantage. They are essentially stealing from everyone else who doesn’t control 25% of ETH hashing power. But, their software is stupid, and took a shortcut.
The software Ethermine is using simulates each transaction and then reads the logs to analyze transaction profitability. They could have chosen to analyze the “opcodes” instead, but that is technically complex, and Ethermine was in a hurry to steal. So this guy created a coin that emits fake transaction events that are recorded in the logs, specifically a fake event that made Ethermine and others like them believe that they were being transferred X number of coins, when they were actually being transferred 0.1X. This led to Ethermine and several others unintentionally donating over 100 ETH to him before they figured out what was happening.
Reacting to the first paragraphs: on decentralized exchanges, traders (liquidity takers) don't change the price because of something called the constant product formula. Market makers (liquidity providers) do.
Since trading doesn't impact price, front-running can't be done by sandwiching buy/sell orders but with LP orders.
Also slippage isn't price movement due your own trading (that's price impact). Slippage is difference between displayed price the moment your press trade and price when order is processed at exchange (due to other people's trading during network delay).
60 comments
[ 3.0 ms ] story [ 110 ms ] threadIn blockchain, the miners control the order of transactions. They do all manner of things to extract profit from your trade because of that control. The simple solution, which is good enough for most people, is to restrict the slippage and assume the miners will get as close to that number as possible. But people are thinking of this problem very hard and ways to solve it.
However, it's not front running legally, because the transactions are public.
HFTs are again, not able to effect your trade. If you place an order in an exchange then HFT's will respond once you have traded, but your trade is guaranteed to go into the market and execute without anyone interfering before hand. If there was liquidity in the market when you tried to trade at the price you entered, you will get that trade. Now it may be that an HFT sees your trade in Chicago and goes off and trades on some other exchange super quickly based on the new information or they might cancel their remaining orders in the makret, but they cannot interfere with your trade before execution.
What's unique about the ethereum situation is that different actors int he market can interfere with your trade before it executes because your desire to trade gets published before it's executed and the miners (in this case I guess sort of analogous to the exchange) are willing to prevent you accessing the market due to a bribe from a third party based on the information you were forced to publish. Which means you never have confidence that you can actually execute a transaction you beleive is available in the market.
Payment for order flow (at least in the case of retail brokerages) is more about the big players preference for trading with unsophisticated players. The rubes just lose more often without any malfeasance required on the side of the brokerage
The energy is not wasted, it is ordered, not demanded and thus never produced. If it were produced it would actually be extremely difficult to get rid of this much energy. You can't just easily dump energy.
As the author is waving their credentials around, I am an active researcher on renewable power grids. It's a bit scary to think that someone with such a basic misunderstanding of the power grid was asked to referee a paper on this, but then a lot of marginal research is published everyday and referees have to come from somewhere...
"For a coal fired power station, for instance, about 2/3 of the energy released when the coal is burnt is discarded as heat in the environment. This reject energy sometimes appears as clouds of vapour coming off a power-station’s cooling towers, such as the well-known ones at Didcot in England."
-- https://energycultures.org/2014/07/rejected-energy-much-ener...
As others have pointed out, the power grid does not even have the capability to "burn off" excess power. It is simply nerver produced in the first place.
Edit: Ok I read the note at the bottom and the graph literally adjusts the rejected energy of renewables to match thermal power plants. Assuming 33% efficiency that means 12.56 phantom units of energy that never even existed in the first place are being rejected.
The only way Bitcoin would use this energy is if they had a contract with the power company to be an on-demand load, switching on and off to help smooth the peaks.
If they aren’t they’re just driving up power use and forcing more operation of less efficient reserve generation on peak.
I understand that a lot of the servers are close to large hydro plants. Under-utilised hydro plants can simply spill water - they spin up and down rapidly compared to thermal plants.
But that is exactly how power plants are using Bitcoin. That's the idea behind Seetee.
I didn't know about Seetee but potentially there are some minor opportunities to mine bitcoins (effectively as an on-demand load) to help balance the grid at short time scales. It also makes sense for a Norwegian Energy company to do this. However this is a tiny amount, especially compared to the efficiency losses the author is presenting in the graphic. I looked up some data I had handy right now and I think we are looking at a few percent of the overall energy consumed at most.
- The visualization used by the author shows US energy consumption while ~65% of mining takes place in China[0] where burning coal remains the dominant energy source.[1]
- Miners run 24/7/365, they add to the baseload!
--
[0] - https://www.statista.com/statistics/1200477/bitcoin-mining-b...
[1] - https://www.eia.gov/international/analysis/country/CHN
98 % of the electricity in Iran is generated from burning either gas or oil. https://www.eia.gov/international/analysis/country/IRN
What surprised me is that "Transportation" in that graph is supposed to "reject" over 22% of its energy, which seems odd to me. For electricity I get it, it's hard to store so if you have excess production there's not much you can do. If your wind turbine can't dump its energy anywhere it's just lost, usually. But for petroleum can't you just... not use it?
As such I wonder if "rejected" means all losses? In which case it's unclear how bitcoin would fit into that.
In general I think the author doesn't make a good enough case. He'd have to show not only that there's enough loss to power the bitcoin network, but also that bitcoin actually slots into these losses instead of adding to the demand.
Actual numbers for coal, oil, and nuclear plants run lower, maybe 40%. The most efficient natural gas plants come close, around 60%, by running at higher temperatures and avoiding conversion to steam.[1]
[1] https://www.brighthubengineering.com/power-plants/72369-comp...
It's a really weird take if that's what the author is saying.
>If Bitcoin used only wasted energy it would consume >2/3rds of the energy of the planet.
>This is why hand-wringing about Bitcoin’s energy use is so frustrating. Bitcoin is pointing at the moon and people are yelling at its finger. The self-congratulatory ignorance I see in this debate is unreal.
I don't get what it means or what point the author tries to make here.
The wasted energy he is talking about here is mostly waste heat and other losses, it is not energy that is actually available for use. That makes the argument pretty silly, since increasing power consumption incurs the exact same losses.
Won't Bitcoin mining show up as part of use used residential and commercial electricity?
And, wouldn't "less coal and natural gas burned" be more desirable than "using that to mine Bitcoin" and sneering about "handwringing"?
But to answer your question: IC engines are energetically inefficient - at best a large fraction of the input potential energy escapes as heat, and actual car engines are worse than ideal.
Transportation is inefficient because ICE engines need to run at lots of different RPM ranges. Hybrid cars have more efficient engines because the engine can run at the most efficient RPM. EVs only have charging and battery losses.
Another, earlier application of the Dark Forest metaphor to ETH / Uniswap by different (?) authors.
The ‘generalized front runner’ described in ^^ seems to me far more pernicious and terrifying than the behavior described in TFA.
The software Ethermine is using simulates each transaction and then reads the logs to analyze transaction profitability. They could have chosen to analyze the “opcodes” instead, but that is technically complex, and Ethermine was in a hurry to steal. So this guy created a coin that emits fake transaction events that are recorded in the logs, specifically a fake event that made Ethermine and others like them believe that they were being transferred X number of coins, when they were actually being transferred 0.1X. This led to Ethermine and several others unintentionally donating over 100 ETH to him before they figured out what was happening.
[1] https://github.com/Defi-Cartel/salmonella
Since trading doesn't impact price, front-running can't be done by sandwiching buy/sell orders but with LP orders.
Also slippage isn't price movement due your own trading (that's price impact). Slippage is difference between displayed price the moment your press trade and price when order is processed at exchange (due to other people's trading during network delay).