price feeding review

[johnson]

算法里的premium 怎么算出来的？是鼓鼓的充值提现费率吗？

premium =Pcex/Pdex-1

这个和充值费率没关系啊。而且也不是负反馈啊。就是内盘跟随外盘联动而已。把内盘和外绑定在一起。如果外盘动，内盘跟随。如果内盘动，就把喂价向相反方向操作。
总的效果就是把内盘活力压制了。导致内盘交易积极性下降。
这个算法和你当初号召喂价改革的初衷不同啊。当时你说内盘交易量大，应该让喂价回归内盘，让内盘有定价权。
但是你仔细看看这个算法，定价权完全交给外盘了

还有一点，你希望bitcny 锚定精准。但是这个喂价公式里没有bitcny 相关的东西。
而且锚定精准与否要看承兑商的手续费计算公式。那个算法你知道么？我比较担心的是，承兑商的算法
有问题，导致喂价算法去适应它，这样看似锚定精准了，但是却是在一个非正常的平衡点

[bitcrab]

算法里的premium 怎么算出来的？是鼓鼓的充值提现费率吗？

premium =Pcex/Pdex-1

[yamtt]

支持巨蟹

[ljk424]

support bitcrab

[bitcrab]

now as I know at least gdex, abit, btspp, magicwallet has adopt "discount tolerance" (other name for "protect discount limit" and I think it is better).

it means that while the smartcoin discount is low enough, price feeding program dot not reduce feed price.

no need to treat premium as same as discount, it's ok not to tolerate premium or just set a little limit for tolerance.

gdex now adopt a limit of -0.5%-0.3%, we plan to upgraded it to -1%-0.3% next time.

the "discount tolerance" can give the market some space, if there's 0 tolerance, DEX price have to adapt closely following CEX price and give DEX users little space.

hope more witnesses can adopt "discount tolerance" method, not only in CNY, but also in USD price feeding.   

[johnson]

算法里的premium 怎么算出来的？是鼓鼓的充值提现费率吗？

[binggo]

这个算法有个问题，就是起始的时候，那个喂价如果是偏离理想值的，那么就会导致经过很长的时间也调整不到理想值。

算法里，至少要考虑一下内盘的价格。完全依赖充值费率这个间接因素有很多弊端。

尤其是当市场不景气，充值提现的人都很少的时候，充值费率就更不能保证客观了。

如果你坚持认为当前的充值费率就代表了市场对bitcny的需求，那么，按照当前的喂价，就应该去打压抵押单，减少bitcny
的产量。就应该有强清，你为什么又要取消强清呢？

我们的目的是什么？维持bitcny的锚定精准？要锚定精准，就要动态的调节bitcny产量， 那该强清的时候就强清。但是，这么一来，
肯定打压了人们参与bts的积极性。
所以，我们的目标不仅仅是维持bitcny的锚定精准，更是要维护bts的发展。

所以，当喂价过低的时候，就应该把喂价提高，bitcny折价到1%有什么关系？5%有什么关系？吸引人们进入bts系统，这才是最关键的。

margin call已经可以起到打压的作用了，根本用不着强清，强清就是个鸡肋，还是动不动就惹麻烦的鸡肋。

不可能放弃锚定精准这一目标。

不可能放弃锚定精准这一目标，

If we can call this is market manipulation？it really is.

[bitcrab]

这个算法有个问题，就是起始的时候，那个喂价如果是偏离理想值的，那么就会导致经过很长的时间也调整不到理想值。

算法里，至少要考虑一下内盘的价格。完全依赖充值费率这个间接因素有很多弊端。

尤其是当市场不景气，充值提现的人都很少的时候，充值费率就更不能保证客观了。

如果你坚持认为当前的充值费率就代表了市场对bitcny的需求，那么，按照当前的喂价，就应该去打压抵押单，减少bitcny
的产量。就应该有强清，你为什么又要取消强清呢？

我们的目的是什么？维持bitcny的锚定精准？要锚定精准，就要动态的调节bitcny产量， 那该强清的时候就强清。但是，这么一来，
肯定打压了人们参与bts的积极性。
所以，我们的目标不仅仅是维持bitcny的锚定精准，更是要维护bts的发展。

所以，当喂价过低的时候，就应该把喂价提高，bitcny折价到1%有什么关系？5%有什么关系？吸引人们进入bts系统，这才是最关键的。

margin call已经可以起到打压的作用了，根本用不着强清，强清就是个鸡肋，还是动不动就惹麻烦的鸡肋。

不可能放弃锚定精准这一目标。

[johnson]

这个算法有个问题，就是起始的时候，那个喂价如果是偏离理想值的，那么就会导致经过很长的时间也调整不到理想值。

算法里，至少要考虑一下内盘的价格。完全依赖充值费率这个间接因素有很多弊端。

尤其是当市场不景气，充值提现的人都很少的时候，充值费率就更不能保证客观了。

如果你坚持认为当前的充值费率就代表了市场对bitcny的需求，那么，按照当前的喂价，就应该去打压抵押单，减少bitcny
的产量。就应该有强清，你为什么又要取消强清呢？

我们的目的是什么？维持bitcny的锚定精准？要锚定精准，就要动态的调节bitcny产量， 那该强清的时候就强清。但是，这么一来，
肯定打压了人们参与bts的积极性。
所以，我们的目标不仅仅是维持bitcny的锚定精准，更是要维护bts的发展。

所以，当喂价过低的时候，就应该把喂价提高，bitcny折价到1%有什么关系？5%有什么关系？吸引人们进入bts系统，这才是最关键的。

[johnson]

0.5% is not enough.  at least 1%

[sschiessl]

price feeding become more critical after the BSIP42 implementation.

currently the feed price in CNY is 4.5% lower than the latest price, in the past several days the gap was even bigger.


according to the negative feedback logic, it's possible and acceptable that the feed price be lower than the latest price, but we still need to review whether there are some flaws here that may hurt the ecosystem.

there are always some price gap between DEX and CEX, if the gap is not big enough, we can not expect the gap be removed by arbitrage，if the CEX price keeps a little lower than DEX price, the negative feedback algorithm may respond and reduce the feed price, in some extreme situation the reducing feed price may trigger continuous margin call which is not so necessary.

while reducing feed price directly lead to margin call that force sell, increasing feed price do not lead to force buy, witnesses need to be more careful while reducing the feed price.

I suggest that we can introduce a "protect discount limit" in the price feeding algorithm, if the smartcoin discount is not bigger than this limit, algorithm does not start the negative feedback process.

that's why gdex-witness updated the algorithm as below:

Code: [Select]

Pdex：BTS price in DEX in smartcoin
Pf: current feed price
premium: current premium

scale=1;
get Pdex, Pf, premium;

while True:
   
   get Pdex, Pf, premium;
   if 0.3%>premium>-0.5%: ##just adopt the current median if the absolute premium is low enough.
       feed price = Pf;
   else:
       feed price = Pf*(1+premium*scale);
   time.sleep(120); ##update every 2 minutes.
I also suggest that each witness publish the price feed algorithm, which can help proxies to review and evaluate the witness work.

The ''protect discount limit' is -0.5% is that algorithm?

[binggo]

I suggest to choose the highest price as the last feed price  from the CEX price and the dynamically feed price.

https://bitsharestalk.org/index.php?topic=27357.0

[bitcrab]

price feeding become more critical after the BSIP42 implementation.

currently the feed price in CNY is 4.5% lower than the latest price, in the past several days the gap was even bigger.


according to the negative feedback logic, it's possible and acceptable that the feed price be lower than the latest price, but we still need to review whether there are some flaws here that may hurt the ecosystem.

there are always some price gap between DEX and CEX, if the gap is not big enough, we can not expect the gap be removed by arbitrage，if the CEX price keeps a little lower than DEX price, the negative feedback algorithm may respond and reduce the feed price, in some extreme situation the reducing feed price may trigger continuous margin call which is not so necessary.

while reducing feed price directly lead to margin call that force sell, increasing feed price do not lead to force buy, witnesses need to be more careful while reducing the feed price.

I suggest that we can introduce a "protect discount limit" in the price feeding algorithm, if the smartcoin discount is not bigger than this limit, algorithm does not start the negative feedback process.

that's why gdex-witness updated the algorithm as below:

Code: [Select]

Pdex：BTS price in DEX in smartcoin
Pf: current feed price
premium: current premium

scale=1;
get Pdex, Pf, premium;

while True:
   
   get Pdex, Pf, premium;
   if 0.3%>premium>-0.5%: ##just adopt the current median if the absolute premium is low enough.
       feed price = Pf;
   else:
       feed price = Pf*(1+premium*scale);
   time.sleep(120); ##update every 2 minutes.
I also suggest that each witness publish the price feed algorithm, which can help proxies to review and evaluate the witness work.