Alex Simpson
@AlexSimpson_96
Followers
1,140
Following
741
Media
46
Statuses
504
Machine Learning research at Tsinghua University. Building MoBuddy, a language tutor in your pocket. Try it free here:
Miami, FL
Joined October 2016
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
Al Smith
• 301793 Tweets
Monsalve
• 232247 Tweets
Catholic
• 129858 Tweets
Watch SA TrueStory YouTube
• 97964 Tweets
Anton
• 83367 Tweets
Boric
• 76845 Tweets
MINQ IN LOVE
• 55151 Tweets
日本シリーズ
• 52400 Tweets
#الهلال_الفيحاء
• 51861 Tweets
#bebekkatilleri
• 51729 Tweets
#ドラゴンボールDAIMA
• 25168 Tweets
Jesus is Lord
• 21942 Tweets
危険運転
• 19329 Tweets
AFFAIR HAPPY ENDING
• 19094 Tweets
Ginés
• 18377 Tweets
TXT OUR STARRY KNIGHTS
• 16910 Tweets
@lightxvision
First year of uni realized I needed social skills to talk to girls. Forced myself to do one thing outside of my comfort zone every day (like go to party, talk to new person etc). By second year had 0 social anxiety and suddenly socializing was super fun.
4
14
2K
1/
@THORChain
are considering dropping ERC-20 tokens from the platform due to the recent exploits.
I was curious how important ERC-20s are to the platform, so I ran the numbers.
Like many, my opinion is ERC-20s should be temporarily paused until fully tested.
12
25
202
1/ When traders on
@THORChain
are fearful, they are willing to pay much higher slip fees. This is great news for LPs and their profits.
More generally, market crashes are some of the most profitable times for THORChain LPs.
7
23
150
1/ In these early days of
@THORChain
, the key drivers of high APYs are not fee revenues. They are Block Rewards and the Incentive Pendulum.
6
21
136
@blader
Taken many machine learning classes at Tsinghua Uni in Beijing and never seen anything like this. Possibly some private business restrictions I’m unaware of, but academic research is thriving as far as I can see. Some really cool chatGPT projects being built at Tsinghua rn.
3
3
126
1/ How much liquidity is enough?
When markets are stable, as little as 1.9 million USD in pool depth can protect users from high slip fees on
@THORChain
.
When the markets are in periods of high greed or fear, even 20 million in depth might not be enough.
6
21
104
1/ Diving into Pool Depths and Composition of Liquidity on
@THORChain
. Two major trends stand out in data:
First, LPs are observant, and employ evolving LP strategies over time. They don't just dump funds and forget about it.
5
19
108
Confused about $XRUNE? I break down its functions, tokenomics and how to purchase it in my latest THORSTARTER 101 guide!
@thorstarter
@runebase_org
4
21
93
1/ LPers on
@THORChain
strongly avoid withdrawing in periods of high Impermanent Loss (IL), but instead are timing their withdrawals to minimise IL, and hence IL Protection (ILP).
This is great news for the Thorchain treasury, since liability is much lower than it could be.
3
10
75
@THORmaximalist
@THORChain
We should try to make some super simplified tutorials to help Doge coin holders move their funds to Thorchain. I’m sure a lot of them will be new to Defi, so a simplified Doge guide filled with memes could help a lot!
Happy to help make that content.
7
4
67
Currently backpacking through South Korea.
Met this fluffy boi at my hostel. His name? THOR.
I think it’s a sign. Time to ape into $RUNE, we going to $100.
3
4
59
10/ The ability to swap native BTC and ETH alone is a huge win for the crypto space, which we shouldn’t forget in these tough times.
Loss of ERC-20 is a hit to the platform, but far from a killing blow.
3
1
48
1
0
47
Teaming up with
@thorstarter
and
@runebase_org
to bring you Thorstarter 101, a series of guides to get you from beginner to participating in IDOs as quickly as possible!
First guide starts it off by answering the question "What is Thorstarter?":
0
4
36
8/ There are other coins yet to be added to Thorchain that could also be a huge win.
@cosmos
tokens e.g. $ATOM have tiny fees, so should behave similarly to BEB2 pools. Lots of arb, lots of volume, and hence lots of profit for LPs.
1
0
36
Amazing resource for anyone looking to access THORChain data, super excited for this!
If you’re a student, start solving bounties in-between classes and you’ll never have to eat instant ramen again.
Calling all THORChain Data Analysts! ⚡️
You’ve been asking for easier data access - the wait is almost over.
@flipsidecrypto
has been hard at work making Midgard data easily accessible.
Here is a preview of what is coming from Flipside Crypto (still in beta)🏗️:
3
14
94
1
4
34
Short explainer of the
@thorstarter
and Olympus Pro bonds program!
Great for users since can purchase XRUNE at discount from market rate.
Great for Thorstarter since gaining ownership of liquidity. Nice alternative to liquidity mining.
1
8
25
4/ This is because of arbitrage opportunities. BEP2 fees much lower, so easier for arbitragers to find profitable swaps. This means more volume, and hence more fees.
Volume on BEP2 is an astonishing 6x higher than ERC-20, at 63 million Rune vs 10 million Rune since launch.
2
1
29
9/ Finally, Thorchain is still a BTC and ETH dominated platform.
Native BTC and ETH pools contribute 37% of total liquidity together (45% before $XRUNE), which is more than all ERC-20s combined.
1
1
25
7/ Removal of ERC-20 might provide new and better opportunities to LPs.
Alternative to USDT could be BEP2 USDT, suggested by
@bullishtrader
on
@THORChainLPU
discord. Low fees, high arb, so APY likely higher than current USDT, despite lower volume. Great news for LPs.
1
2
26
@Joel_P_Atkinson
China is a super interesting country, but reality is it’s not set up well for western tourists. Mostly due to payment issues and Google maps banned. Hope this changes in future though.
3
1
26
6/ Similarly, comparing stable coin pools, BUSD is the clear winner.
BUSD has more fee revenue than USDC and USDT combined, at 56K $Rune vs 45K $Rune. In fact, BUSD is the best performing pool on the whole platform, generating more revenue than even native BTC or ETH.
1
1
24
7/ Unsurprisingly, the key driver of slip fees is pool depth.
BTC, the deepest pool on THORChain, has an average slip fee of 0.13% across all transactions since launch of MCCN.
Raze, the shallowest pool, as an average slip of > 2.1%.
1
2
24
3/ Despite accounting for 35% of all liquidity, ERC-20s account for just 28% of fees earned for LPs. Meanwhile BEP2s account for 25% of total liquidity, but over 40% of fees earned.
1
1
23
14/ 534,278 Rune to LPs, while just 424,736 to nodes.
Inefficient state of the Incentive Pendulum has contributed an additional 214,606 Rune to LPs.
This is a large part of why APYs have been so high.
You can thank mega chads
@THORmaximalist
and
@TheRuneRanger
for that one!
2
0
21
2/ Hello my fellow Thorchads and chadettes! MCCN's return is on the horizon, so I have been diving into the data to see what I can uncover! This Data Digest dives deep into fee revenues, and sees how the slip-based model plays out in practice.
2
3
22
22/ Let's visualise this. If you made it through the math, I'm proud of you 🎉
Here UST price is stable around equilibrium of 1.1 bn UST. As pool balance approaches 0 and 2.25 bn respectively, curve exhibits xyk behaviour, asymptotically deviating from peg to preserve liquidity.
1
1
22
16/ As little as 1.9 mill USD liquidity might be sufficient to keep slip fees low on TC pools in periods of market stability, but an excess of 20 million is going to be necessary to keep fees low across all market conditions.
Another reason to look forward to
#RAISETHECAPS
!
1
1
21
@tsarnick
In RLHF sometimes the humans will be lazy and not write a complete response if a task is too challenging. The model mimics this laziness from the humans.
1
0
21
2/ ERC-20 tokens are important, but less so than they might appear.
ERC-20s account for 35% of all liquidity on the platform. Biggest loss is $XRUNE, USDC and USDT pools, the 2nd, 6th and 8th largest pools respectively.
7
0
20
15/ Unsurprisingly, $XRUNE big winner due to IDO. While cannot expect similar 66% share of liquidity in further cap raises,
@TehSlaw
's tweets make me more bullish than ever on
@thorstarter
, so my feeling is will capture similar liquidity depth to BTC and ETH when ERC-20s restart.
1
1
18
5/ Easy to spot example of this is compare BNB.ETH pool to native ETH. Shortly before the hacks, BNB.ETH APY was at 47%, while native ETH was just 28%.
1
0
19
3/ Second is constant sum. This has the advantage of zero slippage, but can run out of liquidity.
Lets see how that works with some maths (I’ll break it down as simply as I can, don’t worry), and use our equations to do some simulations of depeg scenarios.
1
0
19
@sweatystartup
Having experienced both, it’s very clear to me that entrepreneurship is a more exciting and fulfilling life. I guess depends on the individual.
1
0
16
3/ What's up my Thorchads and chadettes! Full Data Digest can be found here with graphs, tables and all that good stuff:
Here is a thread of the key points.
1
0
16
17/ High opportunity cost to pool ETH, so highly productive BEP2 pools eat into its liquidity.
BTC by contrast has tiny opportunity cost. One of the least productive assets in Defi, so 15%+ APY is immense value.
I expect BTC's important role on THORChain to stay untouched.
2
1
15
12/ Logic of Curve’s stableswap design is to combine these two equations.
When deep liquidity, use constant sum so prices are close to 1:1. When liquidity imbalanced, switch to x*y = k to maintain liquidity.
1
2
18
2/ First, there are two key designs of liquidity pools we need to know about: constant product and constant sum.
Constant product is the most widely used. Might know it from
@THORChain
or
@Uniswap
. Has advantage of continuous liquidity, but can cause high slippage on swaps.
2
0
15
18/ (even coming from a maths background myself I think the equations here get a bit crazy. If that's not your thing, feel free to skip to 22 where I have a visualisation)
For our 4 assets, UST, FRAX, USDT and USDC, we get:
1
1
16
4/ What's up my
#Thorchad
's and chadettes! In this Data Digest I'm diving into Block Rewards and the Incentive Pendulum, which are the keys to high APYs in these early days of MCCN.
Full medium article is here:
Here is a thread of the highlights.
2
0
15
3/ The Incentive Pendulum is also massively inflating rewards to LPs.
Bond wars + limited caps raises consistently creates overbonded state, boosting APYs to LPs.
If you wondered how you were getting such high APYs on pools with tiny trading volumes, this is how.
1
0
15
3/ Hello my fellow Thorchads and chadettes, I'm Alex and I've been diving into
@THORChain
's data to see what we can learn! This week's Data Digest looks at how Impermanent Loss Protection is playing out in practice:
1
1
14
2/ Block Rewards currently make up 75% of income to LPs and Node Operators.
721,145 $Rune was paid out in Block Rewards from launch of MCCN until the first halt, while just 237,869 Rune was earned in fee revenues.
1
0
13
4/ Average pool on Thorchain has 256 unique LPs, with $44,787 average deposit.
Unsurprisingly, BTC and ETH see most widespread demand from LPs, with 1386 and 1141 unique LPs resp. Small average deposits of $21, 870 and $17, 665, suggests some of least whale dominated pools.
1
0
12
17/ Since this proposed curve 4pool has 4 assets, we must expand this equation to 4 assets.
After some rearranging, the generalised Curve Stableswap Equation is as follows:
1
1
14
13/ To my surprise, we see a very different story playing out in practice.
Left chart shows earnings split in efficient state. Right chart shows the actual earnings split.
Instead of 67% to 33% split nodes to LPs, we see almost the opposite. 44% to 56% in favour of LPs.
1
1
13
16/ One big surprise is ETH fell from consistent number 2 spot in liquidity growth to 5th.
ETH overtaken by XRUNE, BEP2 BTC and BUSD.
By contrast, BTC untouched in number 1 spot.
This speaks to what LPs value on THORChain.
2
0
12
10/ To put this into perspective, Uniswap went from 24,963 users to 57,976 in July alone.
Uniswap is only single chain, and still saw 132% user growth.
10k user growth for THORChain is easily within reach.
1
0
12
14/ Using these Time Adjusted data points, we see 880 Rune paid when the market is greedy, 1971 Rune when neutral, and 16,155 Rune when fearful.
The trend is the same. Fear accelerates ILP payouts.
0
0
12
21/ After some crazy calculus we get our price function:
1
0
13
13/ Amazing thing for UST peg is it requires enormous sell pressure to push past its peg.
Let’s do another simulation to see this.
Maths here is more tricky, but logic is same as THORChain example. Differentiate for price function, see what happens for $200,000 sell pressure.
1
0
13
23/ Simulating the same swap size of $200,000 we did for the THORChain pool, we get:
Price=0.99999994
This means UST has moved off peg by 1-0.99999994 = 0.00000006 USD.
0
0
13
19/ To simulate this, I pull current numbers from Curve 3pool. Good chance 4pool attracts deeper liquidity, but this gives initial estimate.
Average liquidity of assets in 3pool is 1.1bn. Therefore I assume the same average liquidity here and set D = 1.1+1.1+1.1+1.1 = 4.4 bn.
1
0
13
2/ Second, while BTC consistently draws a lot of liquidity on each cap raise, ETH and ERC-20s are falling behind over time.
This speaks to what LPs value on Thorchain.
1
0
11
20/ I choose amplification factor A = 100. I fix FRAX and USDC = 1.1 bn for simplicity, and consider a swap of USDT for UST.
Substituting in these values and rearranging our equation for USDT balance we get:
1
0
13
@buchmanster
Even when ETH goes to proof of stake, it will still be using trivial proof of work blocks contained within proof of stake blocks (this makes merge from PoW to PoS easier). This could probably get it past the regulation.
1
0
12
4/ For a two asset liquidity pool, constant sum pool always sets the total quantity of assets in the pool equal to a constant.
This means:
x + y = C
1
0
12
7/ Next we have constant product pools, used by THORChain, Uniswap and many other protocols. These use the equation:
x*y = k
To get our price function, we differentiate, as before:
|dy/dx|= k/x^2
1
0
12
11/ Setting dy/dx equal to this price and solving our system of equations we get:
|dy/dx|=0.98/11.5
x=√(11.5/0.98) k =19,718,100
UST balance in pool increased by:
19,718,100-19,519,950=$198,150
User needs to swap $198,150 UST to set it off peg at a price of 0.98 USD.
1
1
12
14/ Zooming in on final caps raise, can get our best glimpse of what most popular pools will be in a more mature THORChain ecosystem.
Total of 3,598,907 $RUNE was added on final caps raise. Breakdown is in pie chart.
1
0
10
6/ Quick way to see this is take the derivative. Dy/dx gives us the change in y over change in x, or how much y you can get for your x. This is the price!
Here |dy/dx| = 1 meaning the price of USDT is constant at 1 UST.
1
0
12
15/ Want control over how “strong” constant sum part of the equation is, so we add a parameter χ:
(x*y)+ χ(x+y)=D+ (D/2)^2
Want χ to adapt to the ratio of assets in the pool. We make it a function of x and y, and add a constant A for amplification. Hence:
χ=(Ax*y)/(D/2)^2
1
1
12
14/ The curve equation combines the constant sum and constant product equations, so we have:
(x*y)+(x+y)=D+ (D/2)^2
where D is the sum of quantities of asset in a pool.
1
0
12
7/ Does this mean APYs will continually drop each year as block rewards fall off?
Probably not.
Using current MCCN data, we can get a fair estimate of goals for THORChain network to maintain current APYs:
1
0
11
8/ Credit to
@THORChainLPU
and
@GrassRootsio
for this one. This tactical behaviour in LPs shows the effectiveness of education material in the Thorchain space! Investors have clearly read up on IL, and actively try to minimise it.
3
1
10
9/ Pulling in numbers from THORChain’s UST/Rune pool, we can do a quick simulation to see how much sell pressure of UST will cause it to fall off peg to 0.98 USD.
y = Rune quantity = 1,697,387, x = UST quantity = 19,519,960, Rune price is $11.5 USD. k = 33,132,926,344,520.
1
0
11
8/ Key thing to understand here is price is non-constant. It's a decreasing function of x.
As x increases, dy/dx decreases exponentially.
Intuitively, as pool grows imbalanced, prices adjust. This allows liquidity to be continuous, since increasingly expensive to drain pool.
1
0
11
12/ Optimal state of THORChain is when bonded capital matches pooled capital.
67% of system income goes to Nodes, and 33% to LPs. Game theory of the incentive pendulum should create an equilibrium at this efficient state, and this is what I expected to see in the data.
1
1
10
2/ Also, ILP paid out in practice does not correlate 1:1 with IL suffered.
Instead, when the market is booming and greedy, ILP paid out is minimal, as LPers are choosing to stay in pools and accrue fees. However, during periods of high fear, ILP paid massively accelerates.
1
0
10
5/ For example, if you have 1000 UST and 1000 USDT in a pool, C = 2000. When a user swaps in 100 UST, the pool solves:
(1000+100)+ USDT = 2000
USDT = 900
There is 900 USDT left in the pool, meaning the user has received 1000-900 = 100 USDT. Price of the USDT is 100/100 = 1.
1
0
10
3/ A total of 237,868 $Rune paid out in fees from MCCN launch to first system halt. BUSD is top earner.
BTC and ETH are also high earners, but less so than might expect. Together are 37% of liquidity, but just 28% of fees earned.
This plot shows fees earned for all pools:
1
2
9
13/ Many ERC-20 pools e.g. $HOT, $SNX, see large initial spike of liquidity, but little change later on.
They see significant seed funding to get listed, but fail to capture additional liquidity from LPs.
Low fee revenues + high opportunity cost makes them unattractive to LPs.
1
0
10
16/ Now as pool gets increasingly unbalanced, χ decreases and equation is more like xky. As a quick demonstration of this:
Balanced pool:
χ=(A*100*100)/(200/2)^2 =A
Unbalanced pool:
χ=(A*5*100)/(105/2)^2 =0.18A
χ decreased from A to 0.18A as pool became unbalanced.
1
0
10
11/ The other key to high APYs is the Incentive Pendulum.
Bond Wars, coupled with limited caps raises to add liquidity, is consistently leaving THORChain overbonded.
This is great news for LPs. It massively boosts their APYs.
1
0
9
@levelsio
China is cool but it’s hard to recommend as a tourist destination. Firewall is difficult to adapt to. Credit cards don’t work in most places. Requires a lot of effort to travel there.
1
0
10
4/ BUSD is top performer for two reasons.
First, demand for stable coins was very high in recent market crash. This increases volume, and hence fees earned.
Second, BEP2 pools have low fees. This increases arbitrage opportunities, again increasing volume and fees earned.
1
0
9
2/ What's up my
#Thorchad
's and chadettes! In this Data Digest Report I'm exploring the lower bounds of liquidity necessary to maintain low slip fees on THORChain.
The full report is here, but here is a thread of key points:
1
1
10
@levelsio
Only spent 3 months in Korea but this was my observation too. Been to about 60 countries, Korea seemed the most unhappy, Denmark probably the happiest.
1
1
9
3/ First, why is deep liquidity important on
@THORChain
?
TC uses Continuous Liquidity Pools, where fees are proportional to the share of liquidity a swap demands.
When liquidity is low, this can make swap fees prohibitively high, and push users away from the platform.
1
0
9
10/ When the pool is balanced and UST is on peg, the price of UST in Rune is:
1/11.5
When UST is off peg at 0.98 USD, it has a price in terms of Rune of:
0.98/11.5
1
0
9
@levelsio
Think it highly depends on the country you come from. If you are from the US and need to take crazy debt to study, skip it. If you from countries like Denmark, which has free uni and you get a living wage as a student, go uni and enjoy urself.
0
0
8
9/ This can be achieved by a 53,114,511 Rune increase in volume.
Equivalently, an increase in daily average volume from 1,118,280 to 1,683,329 Rune.
In terms of users, requires an additional 10,043 unique users brought to the platform.
1
0
9
8/ 721,145 Rune emitted in block rewards, 237,868 Rune earned in fees.
A 1/6th decline in block rewards means a 120,191 Rune reduction in earnings.
This will require just over a 50% increase in fee revenues to compensate.
1
0
9
5/ Here we can see the system income on THORChain from launch of MCCN up to the first system halt.
Aside from a spike in fee revenues on day 45 (caused by a market crash), block rewards have consistently outpaced fee revenues since being turned on.
1
0
9
9/ Secondly, the opportunity cost of pooling ERC-20 tokens on THORChain is high. These are some of the most productive assets in Defi, so in an environment of limited cap raises, LPs are choosing to pool other assets instead.
1
0
9
@GrassRootsio
What makes ERC-20s so much harder to integrate than others? How come it’s fine for single chain DEXs like Uniswap, Sushi etc, but really difficult on multi chain?
2
0
8
5/ ERC-20 pools have vastly different liquidity composition.
$DODO, $ALCX among others have very few unique LPs, over $100,000 average deposit, suggesting heavily whale dominated.
To see why this happens, we need to understand the decision making of LPs.
1
0
9
@novemberviolet1
@Joel_P_Atkinson
Nah, lived in China 2019 for Masters. I guess it was more open back in those days tho?
0
1
7
6/ Intuitively, might expect the opposite.
Large transactions put pools more out of balance, leading to higher slip. However, data tells a different story.
Corroborating my
@terraswap_io
findings, we see slip fees dictate transaction sizes, rather than other way around.
1
0
8